Every check, in plain words.

80 rules across your pull requests, Preview deployments and live site. Each says what it catches, why it matters and how to fix it.

On every pull request 27 code checks for Next.js apps. Paradigma reads your source and never runs it.

Indexing

A page turned noindexA page or layout that Google could index now tells search engines not to, for example robots: { index: false } in its metadata.Blockerindexing/noindex-added
Why it matters
Once Google recrawls the page, it drops out of search results. AI answer engines that rely on search indexes, such as Google AI Overviews, ChatGPT search and Perplexity, stop citing it too.
How to fix
Remove index: false (or noindex) from the metadata. If the page really should stay out of search, mute the rule for the repository from the check run.
  • A GitHub suggested change on the pull request, committed in one click.
  • A draft fix pull request, opened from the dashboard.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Severity
Warning when generateMetadata returns noindex on only some of its paths.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute noindex rule” on the check run, for one repository

Search and AI crawlers

Search engines blocked in robots.txtA change to robots.txt or app/robots.ts stops Googlebot or Bingbot from crawling pages they could crawl before.Blockercrawlers/search-engine-blocked
Why it matters
Blocked pages stop being crawled, fall out of search results over time, and can't appear in Google AI Overviews or Copilot answers.
How to fix
Allow the crawlers again, and disallow only the private paths you meant to hide.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Severity
Blocker when the whole site is blocked; warning when only some routes are.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute search bots” on the check run, for one repository
AI search crawlers blockedrobots.txt newly blocks the crawlers AI assistants use to fetch and cite pages: OAI-SearchBot, ChatGPT-User, PerplexityBot, Perplexity-User, Claude-SearchBot and Claude-User.Blockercrawlers/ai-search-blocked
Why it matters
These crawlers fetch pages so AI assistants can cite them. Blocking them can take your site out of ChatGPT, Perplexity and Claude answers, even while you still rank on Google.
How to fix
If you meant to opt out of AI training, block the training crawlers (GPTBot, ClaudeBot, Google-Extended) and keep the search crawlers allowed.
  • A GitHub suggested change on the pull request, committed in one click.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Severity
Blocker when the whole site is blocked; warning when only some routes are.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute AI search bots” on the check run, for one repository
AI training crawlers blockedrobots.txt newly blocks crawlers that collect data for AI training: GPTBot, ClaudeBot, Google-Extended, Applebot-Extended and CCBot.Infocrawlers/ai-training-blocked
Why it matters
Blocking training crawlers is a legitimate choice, and it doesn't remove you from AI search citations, which use separate crawlers such as OAI-SearchBot and Claude-SearchBot. Paradigma reports it so the choice is deliberate.
How to fix
Nothing to fix if the block is intended. Mute the rule once you've decided.
  • No Autopilot fix. Blocking training crawlers is your business choice.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute AI training” on the check run, for one repository

Titles, canonicals and metadata

metadataBase changed or removedmetadataBase in a page or layout is removed, or now points to a different host.Warningmetadata/base-url-changed
Why it matters
Next.js resolves relative canonical, Open Graph and alternate URLs against metadataBase. Without it, Next.js falls back to the deployment URL, so canonicals and shared links can point at a preview or localhost domain, and search engines may consolidate rankings onto the wrong host.
How to fix
Point metadataBase at your production origin.
  • A GitHub suggested change on the pull request, committed in one click.
  • A draft fix pull request, opened from the dashboard.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Severity
Blocker when the new host looks like a preview or local domain (localhost, 127.0.0.1, *.vercel.app, *.netlify.app).
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute metadataBase” on the check run, for one repository
Canonical URL changed or removedA page's alternates.canonical is removed, or now points to another path or another host.Warningmetadata/canonical-changed
Why it matters
Search engines follow the canonical. Pointing it elsewhere asks them to index the other URL instead of this page, which moves its rankings and AI citations. Removing it lets URL variants (query strings, trailing slashes, duplicate paths) compete with each other.
How to fix
Point the canonical back at this page's own URL.
  • A GitHub suggested change on the pull request, committed in one click.
  • A draft fix pull request, opened from the dashboard.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Severity
Blocker when the canonical moves to another host.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute canonical rule” on the check run, for one repository
hreflang alternates removed or brokenalternates.languages is removed, drops a language, or points at a route that doesn't exist in the pull request.Warningmetadata/hreflang-changed
Why it matters
Without hreflang, search engines pick one language version per query and the others compete with it instead of serving their own audience. Pairs that point to a 404 are dropped.
How to fix
Restore the missing languages, or point each alternate at a route that exists.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute hreflang rule” on the check run, for one repository
Page title or description removedA page loses the title or description it set in its metadata, or a layout stops providing one that its pages inherited.Warningmetadata/tags-removed
Why it matters
Without its own title, a page inherits the layout's, so several pages can show the same title in search results and AI answers. Without a description, search engines and AI assistants write their own snippet from page text.
How to fix
Restore the title and description, or write new ones for the page.
  • A GitHub suggested change on the pull request, committed in one click.
  • A draft fix pull request, opened from the dashboard.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Severity
Info when only the description is removed.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute title/desc rule” on the check run, for one repository
Duplicate page titlesA pull request gives a page the same title as another page under the same layout.Warningmetadata/duplicate-title
Why it matters
When two pages share a title, search results and AI citations can't tell them apart, and the wrong page may be shown.
How to fix
Give each page a title that names what is on it.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute dup titles” on the check run, for one repository

Structured data

Structured data removedA page or layout that rendered <script type="application/ld+json"> no longer does.Warningstructured-data/removed
Why it matters
Structured data (schema markup) powers rich results and helps search engines and AI assistants understand products, articles, FAQs and organizations.
How to fix
Restore the JSON-LD script. If it moved to another component, mute the finding.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute JSON-LD rule” on the check run, for one repository

Sitemaps

URLs dropped from the sitemapapp/sitemap.ts stops listing URLs it listed before.Warningsitemap/urls-dropped
Why it matters
Crawlers use the sitemap to find and revisit pages. Dropping a URL that still exists slows down how quickly its changes are picked up; if the page is gone, it needs a redirect too.
How to fix
Add the URLs back if the pages still exist; redirect them if they don't.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute sitemap URLs” on the check run, for one repository
Sitemap removedA pull request deletes the sitemap (app/sitemap.ts, app/sitemap.xml or a sitemap*.xml in public/) and adds no replacement.Warningsitemap/removed
Why it matters
Sitemaps are how search engines and AI crawlers discover new and updated pages quickly. Without one, new pages can take much longer to be indexed or cited.
How to fix
Keep a sitemap in the app, or replace the one you removed.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute sitemap rule” on the check run, for one repository

Routes and redirects

A page now redirectsA page that rendered content now calls redirect() or permanentRedirect(), or a permanent redirect becomes a temporary one.Warningroutes/page-now-redirects
Why it matters
redirect() answers with a temporary 307, so search engines keep the old URL indexed and don't pass its rankings on. A permanent 308 moves rankings to the destination over time, so internal links and the sitemap should follow.
How to fix
Use permanentRedirect() if the move is permanent, then update internal links and the sitemap.
  • The Paradigma agent opens a pull request, but route changes are high risk: they always wait for your review.
Severity
Info when the new redirect is permanent.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute redirect rule” on the check run, for one repository
Route removed without a redirectA route that existed before the pull request is gone, and next.config has no redirect for it.Warningroutes/removed-without-redirect
Why it matters
Links, bookmarks and search results pointing to the old URL return 404, which drops its rankings and any AI answers citing it.
How to fix
Add a permanent redirect from the old route to its closest replacement.
  • A GitHub suggested change on the pull request, committed in one click.
  • The Paradigma agent opens a pull request, but route changes are high risk: they always wait for your review.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute removed routes” on the check run, for one repository

Server rendering

Pages that load their content only in the browserA public page.tsx marked "use client" that fetches the content it shows after it loads, in useEffect or with SWR or React Query.Warningnextjs/client-only-content
Why it matters
The server sends HTML without the content, which arrives only after JavaScript runs. Google renders it later, if at all, and AI crawlers like GPTBot and ClaudeBot read the raw HTML, so they see an empty page.
How to fix
Fetch the data in a Server Component (an async page, with generateStaticParams for dynamic routes) and pass it to the interactive part as props.
  • Autopilot can move the page's data loading into a Server Component and keep the interactive parts as small client components, so visitors see the same page. It always waits for your review.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute client-only” on the check run, for one repository

Images

Images missing alt textA new image on a public page with no alt attribute.Infoimages/alt-missing
Why it matters
Alt text is how search engines, image search and AI assistants understand an image, and how screen readers describe it. Without it, the image says nothing to them.
How to fix
Add alt text that says what the image shows.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute image alt” on the check run, for one repository
Image loading priority removedAn image that was marked to load early loses that hint: preload (or the older priority) on next/image, fetchPriority="high" or loading="eager".Infoimages/priority-removed
Why it matters
The hero image is often the Largest Contentful Paint element. Loading it late can slow LCP, one of the Core Web Vitals.
How to fix
If the image is still above the fold, restore an early-loading hint.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute image priority” on the check run, for one repository

Penalty risk

Many new pages in one pull requestA pull request that adds many public page routes, or many sitemap URLs, at once compared with the size of the site.Warningrisk/page-burst
Why it matters
Publishing many pages at once, each with little of its own, is what Google's scaled content abuse policy describes.
How to fix
Ship new pages in smaller steps, each with real substance, and keep thin ones out of the index until they have it.
  • No Autopilot fix. A warning on your own pull request; Autopilot never adds pages.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute page bursts” on the check run, for one repository
New near-identical pagesThree or more pages added in one pull request whose text is mostly the same and differs only by a word or two, such as a city or a keyword.Warningrisk/templated-pages
Why it matters
Pages made from one template for many places or keywords are what Google's doorway policy describes.
How to fix
Merge them into one strong page, or give each page content a visitor from that search needs.
  • No Autopilot fix. A warning on your own pull request; Autopilot never generates text to make pages differ.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute templated pages” on the check run, for one repository
New text hidden from visitorsA long block of text added inside an element styled to be invisible: display none, zero font size, zero opacity or pushed off screen.Warningrisk/hidden-text
Why it matters
Text visitors cannot see but crawlers read is what Google's hidden text policy describes.
How to fix
Show the text, or remove it if it exists only for search engines or AI assistants. Menus, dialogs and accessible labels are fine.
  • No Autopilot fix. On the live site, Autopilot can propose removing it as a draft.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute hidden text” on the check run, for one repository
New instructions to AI assistantsText added to a page, component or attribute that tells AI assistants what to do, such as asking them to drop their earlier guidance or to favour the site in their answers.Warningrisk/ai-instructions
Why it matters
Bing treats text that tries to steer AI answers as abuse, and Google's spam policies cover it too.
How to fix
Remove the instructions. Write for people; assistants read the same page.
  • No Autopilot fix. On the live site, Autopilot can propose removing it as a draft.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute AI instructions” on the check run, for one repository
Redirects for crawlers or phones onlyA change to the proxy, middleware or Next.js config that redirects to another address only when the user agent (a crawler, a phone) or the referrer matches.Warningrisk/crawler-redirect
Why it matters
Sending crawlers, phones or visitors from Google somewhere other than everyone else is what Google's sneaky redirects policy describes.
How to fix
Serve everyone the same page, or make sure crawlers get the same redirect when it is the same site (a mobile host or a country domain).
  • No Autopilot fix. Redirect logic is the owner's call; Autopilot never edits it.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute UA redirects” on the check run, for one repository
Self-serving review markupJSON-LD added in the pull request that puts a rating or review on the site's own Organization, LocalBusiness or WebSite.Warningrisk/review-markup
Why it matters
Google treats ratings an organization gives itself as self-serving and never shows them; misleading markup can lead to a structured data manual action.
How to fix
Remove the rating from your own organization. Keep product ratings only where visitors see the reviews.
  • No Autopilot fix. On the live site, Autopilot can propose removing it as a draft.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute self reviews” on the check run, for one repository
Dates changed without contentA sitemap lastModified or JSON-LD dateModified set to the build or request time, or a file where only the modified dates change.Warningrisk/fake-freshness
Why it matters
Dates that change without the content teach Google to ignore your dates.
How to fix
Take each date from when that content last changed, or leave it out.
  • No Autopilot fix. On the live site, Autopilot can propose fixing the date source as a draft.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute date bumps” on the check run, for one repository
llms.txt addressing AI assistantsLines added to llms.txt or a Markdown version of a page that give AI assistants instructions.Warningrisk/llms-mirror
Why it matters
A version written for AI that says more than the page, or tells assistants what to say, is read as cloaking by AI answer engines. No engine has a written rule for llms.txt, so this is inferred.
How to fix
Keep llms.txt and Markdown versions to what the HTML pages say.
  • No Autopilot fix. On the live site, Autopilot can propose aligning the file as a draft.
Severity
Penalty risk is a warning only: it never blocks the pull request.
Runs
Every pull request, from the source code
Shows up in
The GitHub check on the pull request, pointing at the file
Mute
“Mute llms.txt rule” on the check run, for one repository

On your Preview deployment 13 checks on the pages your Preview actually serves, compared with production.

Rendered pages

Pages that error on the deploymentA public page on your Preview deployment answers with an HTTP error (4xx or 5xx).Blockerrendered/page-error
Why it matters
Search engines and AI crawlers can't read a page that errors. If it ships, the page drops out of results.
How to fix
Fix the error, or redirect the route if it was meant to go away.
  • No Autopilot fix. The cause is usually in the deployment, its data or its environment.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute page errors” on the check run, for one repository
Canonical pointing at the deployment URLThe rendered page declares its own deployment host (*.vercel.app, *.netlify.app, *.pages.dev) as the canonical URL.Blockerrendered/canonical-host
Why it matters
This usually means metadataBase isn't set, so Next.js falls back to VERCEL_URL. Search engines would index the deployment URL instead of your domain.
How to fix
Set metadataBase to your production origin in the root layout.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute canonical host” on the check run, for one repository
Invalid structured dataA <script type="application/ld+json"> block on the rendered page isn't valid JSON.Warningrendered/json-ld-invalid
Why it matters
Search engines ignore structured data that doesn't parse, so the page loses its rich results.
How to fix
Build JSON-LD with JSON.stringify instead of string templates.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute invalid JSON-LD” on the check run, for one repository
Rendered pages without a titleThe HTML your Preview deployment serves has no <title>, even if the code looks fine.Warningrendered/title-missing
Why it matters
Search engines and AI assistants invent a title from page text when there isn't one.
How to fix
Make sure every page resolves a title, with a fallback for missing data.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute missing titles” on the check run, for one repository
Rendered pages without a descriptionThe HTML your Preview deployment serves has a title but no meta description.Inforendered/description-missing
Why it matters
Without a description, the snippet in search results and AI answers is written from page text.
How to fix
Give every public page a description, with a fallback for missing data.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute missing descs” on the check run, for one repository

Preview compared with production

Title removed on the PreviewA page that has a title in production renders without one on the Preview of your pull request.Warningpreview/title-removed
Why it matters
Search results and AI citations may use unrelated page text instead of your title.
How to fix
Restore the title before merging.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute lost title” on the check run, for one repository
Description removed on the PreviewA page that has a meta description in production renders without one on the Preview.Infopreview/description-removed
Why it matters
The search snippet will be written from page text instead.
How to fix
Restore the description before merging.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute lost desc” on the check run, for one repository
Canonical removed on the PreviewA page that declares a canonical URL in production renders without one on the Preview.Warningpreview/canonical-removed
Why it matters
URL variants may compete in search results after deployment.
How to fix
Restore the canonical before merging.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute canonical” on the check run, for one repository
Structured data removed on the PreviewA page that renders valid JSON-LD in production renders none on the Preview.Warningpreview/json-ld-removed
Why it matters
The page can lose its rich results, and search engines and AI assistants lose the structured facts about it.
How to fix
Check whether the structured data moved or was removed, and restore it.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute lost JSON-LD” on the check run, for one repository
H1 removed on the PreviewA page with a top-level heading in production renders without one on the Preview.Infopreview/h1-removed
Why it matters
The H1 is how a page names its main topic for readers, search engines and AI assistants.
How to fix
Keep one H1 that names the page's main topic.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute lost H1” on the check run, for one repository
Content lost on the PreviewA page serves far less main content on the Preview than in production.Warningpreview/content-shrunk
Why it matters
Search engines and AI crawlers read the HTML. If the content now loads only in the browser, or a data fetch fails on the deployment, they see a near-empty page.
How to fix
Render the main content on the server, then check data loading on the deployment.
  • No Autopilot fix. Content loss can be intentional, so a person needs to decide.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute content loss” on the check run, for one repository
New redirect on the PreviewA route that production serves directly (or at a different path) redirects somewhere else on the Preview.Warningpreview/redirect-added
Why it matters
If the redirect ships by mistake, visitors and crawlers land somewhere else, and the page's rankings start moving with them.
How to fix
Limit the redirect to the routes it is meant for, and update the sitemap and internal links if the move is intended.
  • The Paradigma agent opens a pull request, but route changes are high risk: they always wait for your review.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute new redirect” on the check run, for one repository
New penalty risk on the previewA penalty risk check (affiliate links, self-serving ratings, hidden text, instructions to AI assistants, generated-text leftovers, llms.txt claims and others) that passes on production but fails on the preview.Warningpreview/penalty-risk
Why it matters
It shows a risk this pull request would ship before it reaches Google, Bing or AI answers.
How to fix
Follow the fix the warning names, in the same pull request.
  • No Autopilot fix. A warning on your own pull request.
Severity
Never blocks the pull request; a review lead is a note.
Runs
Every pull request, once its Preview deployment is ready
Shows up in
A GitHub check on the pull request
Mute
“Mute preview risk” on the check run, for one repository

On your live site 40 monitoring checks: Search Console, Bing, Core Web Vitals, crawler access and agent discovery.

Google Search Console

Search position dropsA search query where your Google position fell.Warninggsc/query-position-drop
Why it matters
Position drives clicks. A query sliding down the results means fewer people find you through it, and the same pages are often what AI answers draw on.
How to fix
Check the affected page and what changed recently, on your site and in the results for that query.
  • Autopilot can propose a better title and description as a pull request. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Search click dropsA page that gets clearly fewer clicks from Google Search.Warninggsc/page-click-drop
Why it matters
Clicks are your organic traffic. Catching a drop early, and seeing which release lines up with it, is the difference between a quick fix and a lost month.
How to fix
Look at what changed on the page and around it. On paid plans, the change timeline lines the drop up with your releases.
  • Autopilot can propose a better title and description as a pull request. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Low click-through rateA page that gets fewer clicks per impression than your other pages at a similar position.Infogsc/ctr-opportunity
Why it matters
Your title and description are the search result. A page that ranks but doesn't get clicked usually has a snippet that doesn't match what people searched for.
How to fix
Rewrite the title and description so they answer the queries the page shows up for.
  • Autopilot can propose a better title and description as a pull request. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans

Bing Webmaster Tools

Bing click dropsA page that gets clearly fewer clicks from Bing Search.Warningbing/page-click-drop
Why it matters
Bing is a search engine in its own right, and a sudden drop there is often the first sign of an indexing problem.
How to fix
Check indexing, rankings and demand before changing the page.
  • Autopilot can propose a better title and description as a pull request. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans

Core Web Vitals

LCP regressionsYour real-user Largest Contentful Paint (LCP) went from good to not good.Warningperf/crux-lcp-regression
Why it matters
LCP is one of Google's Core Web Vitals: how long the main content takes to appear for real visitors. Slow pages lose visitors before they read anything.
How to fix
Find what slowed the main content (hero image, fonts, server response) and fix that.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
INP regressionsYour real-user Interaction to Next Paint (INP) went from good to not good.Warningperf/crux-inp-regression
Why it matters
INP is the Core Web Vital for responsiveness: how long the page takes to react to a click or keypress.
How to fix
Find the interactions that block the main thread and make them lighter or deferred.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
CLS regressionsYour real-user Cumulative Layout Shift (CLS) went from good to not good.Warningperf/crux-cls-regression
Why it matters
CLS is the Core Web Vital for visual stability: how much the page jumps around while it loads.
How to fix
Reserve space for images, embeds and late content so nothing pushes the page around.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans

Production checks

Live page fails to loadA production page returns an HTTP status outside 200–299 to Paradigma's crawler.Blockerwatch/page-error
Why it matters
A page that errors can drop out of Google and can't be read or cited by AI assistants. It often comes from a deploy, a CMS outage or a hosting rule rather than a code change.
How to fix
Open the page, find what returns the error (deploy, CMS, redirect loop, firewall) and restore it, or redirect it if it moved.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page marked noindexA production page carries noindex in its robots meta tag or X-Robots-Tag header.Blockerwatch/noindex
Why it matters
Noindex removes the page from Google and from AI answers built on search indexes. Staging settings, CMS toggles and proxy rules can add it without any pull request.
How to fix
Remove the noindex from the page or the rule that adds it, unless hiding the page is intended.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Canonical URL points to another domainA production page declares a canonical URL on another origin.Warningwatch/canonical-offsite
Why it matters
An off-site canonical tells Google to index the other site's URL instead of yours. It's common after copying templates or moving domains.
How to fix
Point the canonical at the page's own URL on your domain, unless the content is intentionally syndicated.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page has no titleA production page renders without a <title>.Warningwatch/title-missing
Why it matters
Search results and AI citations show the title. Without one, Google writes its own, often poorly, and the page is harder to choose.
How to fix
Give the page a specific title in its metadata or template.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Structured data that isn't valid JSONA JSON-LD block on a production page can't be parsed as JSON.Warningwatch/json-ld-invalid
Why it matters
Search engines skip structured data they can't parse, so rich results and the facts it describes are lost.
How to fix
Fix the JSON in that block (often a trailing comma, a stray quote or unescaped content from a CMS).
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page lost its canonical URLA production page that declared a canonical URL on the previous crawl no longer declares one.Warningwatch/canonical-removed
Why it matters
Without a canonical, duplicate URLs (parameters, trailing slashes, copies) compete with each other in search.
How to fix
Restore the canonical in the page's metadata or template.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page lost its meta descriptionA production page that had a meta description on the previous crawl no longer has one.Warningwatch/description-removed
Why it matters
Search results often show the description. Losing it lets Google pick a random excerpt, which can lower clicks.
How to fix
Restore the description in the page's metadata or CMS field.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page lost its structured dataA production page that had JSON-LD on the previous crawl has none now.Warningwatch/json-ld-removed
Why it matters
Structured data feeds rich results and gives search engines and AI systems clear facts about the page. Losing it quietly removes both.
How to fix
Restore the JSON-LD that the template or CMS used to output.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page lost its H1A production page that had an H1 on the previous crawl has none now.Infowatch/h1-removed
Why it matters
The H1 tells readers, search engines and AI systems what the page is about. A redesign or CMS change can drop it without anyone noticing.
How to fix
Restore a single, specific H1 on the page.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page lost most of its contentA production page's main content is much shorter than on the previous crawl.Warningwatch/content-shrunk
Why it matters
When a CMS call fails or content moves to client-side rendering, crawlers see an almost empty page and rankings and AI citations follow.
How to fix
Check the CMS response and server rendering for that page and restore the content in the HTML the server returns.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live page started redirectingA production URL that was served directly before now redirects elsewhere.Infowatch/redirect-added
Why it matters
A permanent move is fine, but sitemaps and internal links should point to the new URL so crawlers don't waste requests and signals aren't split.
How to fix
If the move is permanent, update the sitemap and internal links to the new URL; if not, remove the redirect.
  • No Autopilot fix. The cause is often outside the repo (a CMS, CDN or hosting rule), so Paradigma reports it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Key page not updated in over a yearA key page whose own update date is more than 12 months old. Key pages are your home page, top 20 pages by Search Console clicks, pages AI answers cited and top 20 pages by Google AI impressions.Infowatch/stale-key-page
Why it matters
Most pages that AI answers cite were updated in the last year, and what counts is the update date, not the publication date. Old prices, numbers and screenshots on your best pages also cost trust with readers.
How to fix
Review the page's facts, numbers, prices, screenshots and links, and update what's out of date. Change dateModified and the sitemap lastmod only when the content really changes, and keep the URL.
  • Autopilot plans a content refresh: it updates outdated facts it can verify in the repo and never changes only the date. Content kept in a CMS needs a person. It always waits for your review.
Severity
A warning for your top 5 pages by Search Console clicks; a note for the other key pages.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Sitemap lastmod that doesn't track real changesEvery URL in the sitemap shares one <lastmod>, or <lastmod> changes on every crawl while the pages' content doesn't.Warningwatch/lastmod-unreliable
Why it matters
Search engines use lastmod only while it matches real edits. A build time stamped on every URL, or the current time on every request, teaches them to ignore it, so a real update to an important page loses that signal.
How to fix
Set each URL's lastmod from when its content last changed (the CMS or file update time), or leave it out until you have a real date.
  • Autopilot plans a sitemap fix that takes each URL's date from its real update time; the date source is usually one line in the sitemap generator. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
dateModified moved without a content changeA page's dateModified in JSON-LD changed since the previous crawl, while its main text is identical.Infowatch/date-without-change
Why it matters
A date that moves on its own, often a template printing the current time, is fake freshness. Search engines and AI crawlers learn to distrust the site's dates, including real updates.
How to fix
Set dateModified from when the content was last edited, not from the time the page is rendered.
  • No Autopilot fix. The date usually comes from a CMS field or a template shared by many pages, so Paradigma reports it for a person to check.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Crawlers excluded by the live robots.txtYour production robots.txt disallows a search or AI crawler, such as Googlebot or OAI-SearchBot, on a page.Warningwatch/bot-robots-blocked
Why it matters
It catches blocks that never went through a pull request: a CDN rule, a CMS setting or another app serving robots.txt. A page Google can't crawl falls out of search; a page OAI-SearchBot can't crawl can't be cited in ChatGPT search.
How to fix
Find what serves the live robots.txt and allow the crawler again, unless the exclusion is intended.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Crawler user agent gets an HTTP errorA test request with a crawler's user agent, such as Googlebot or OAI-SearchBot, gets an HTTP error while an ordinary request to the same page works.Infowatch/bot-http-denied
Why it matters
Bot protection, firewalls and CDN rules sometimes block crawler user agents. The test doesn't prove the real crawler is blocked, because Paradigma's IP isn't a verified crawler IP, so confirm in your CDN rules and real crawler logs.
How to fix
Check your CDN and bot-protection rules for the crawler, and confirm with real crawler logs or the provider's tools.
  • No Autopilot fix. The fix is in your infrastructure, not your code.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Crawler user agent gets a CDN bot challengeA test request with a search or AI crawler's user agent gets a bot challenge or block page from your CDN or firewall (Cloudflare, Akamai, DataDome, Vercel or a captcha), while an ordinary request to the same page works.Warningwatch/bot-challenge
Why it matters
Since July 2025, Cloudflare blocks AI crawlers by default on new domains, and owners often don't know. A challenged crawler can't read the page, so it can't index or cite it. The test comes from Paradigma's IP, and CDNs verify real crawlers by IP, so the challenge may only be IP verification: the real crawler may still get through.
How to fix
Open your CDN dashboard and check the bot and AI crawler settings (on Cloudflare, AI Crawl Control and Bot Fight Mode). Allow the crawlers you want to be indexed and cited by, then confirm with real crawler logs.
  • No Autopilot fix. The setting lives in your CDN dashboard, not your code.
Severity
Never a blocker: a request from Paradigma's IP can't prove the real crawler is challenged.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Noindex served only to crawlersA crawler user agent, such as Googlebot or OAI-SearchBot, receives noindex on a page that ordinary visitors get without it.Warningwatch/bot-noindex
Why it matters
A noindex served only to crawlers hides the page from search or AI answers while it looks fine in a browser. Confirm with real crawler logs or the provider's tools, because Paradigma's IP isn't a verified crawler IP.
How to fix
Find the rule that treats crawlers differently and remove the noindex from it.
  • No Autopilot fix. The cause is often a CDN or hosting rule, so a person needs to confirm it.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live pages that are empty without JavaScriptA production page whose HTML has almost no text and the markers of a client-rendered app.Warningwatch/js-shell
Why it matters
AI crawlers like GPTBot and ClaudeBot read the HTML the server sends and don't run JavaScript, so a client-rendered page looks empty to them and can't be cited. Google renders JavaScript, but later and not always.
How to fix
Render the main content on the server: server-side rendering or static generation in your framework (in Next.js, a Server Component or generateStaticParams), or prerendering for a single-page app.
  • On Next.js and other frameworks that render on the server, Autopilot can move the page's data loading there; on a single-page app it reports what to change instead of editing. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live pages whose content needs JavaScriptA production page that shows far more main text or same-site links after its scripts run than in the HTML the server sends.Warningwatch/js-only-content
Why it matters
AI crawlers like GPTBot and ClaudeBot read the HTML the server sends and don't run JavaScript. Text they never see can't be cited, and links they never see don't lead them to your other pages. Google renders JavaScript, but later and not always.
How to fix
Render the missing content and navigation on the server: server-side rendering or static generation in your framework (in Next.js, a Server Component or generateStaticParams), or prerendering for a single-page app. Use plain <a href> links for navigation.
  • On Next.js and other frameworks that render on the server, Autopilot can move the page's data loading there; on a single-page app it reports what to change instead of editing. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Many new pages at onceThe sitemap gained many URLs in a short time for a site of its size, or a domain registered only weeks ago already lists a large sitemap. Compared crawl to crawl from hashes of the sitemap's URLs.Warningwatch/page-burst
Why it matters
Google's scaled content policy targets many pages made to rank with little value of their own, and young sites that ship many pages at once are a pattern spam updates have demoted. A burst is a reason to look, not proof of anything.
How to fix
Check that each new page has real substance before it is indexed. Keep thin ones out of the index (noindex and out of the sitemap) or merge them, and add pages in smaller steps.
  • No Autopilot fix. Removing or noindexing pages is the owner's call; Autopilot never creates or removes pages for this.
Severity
Shown as penalty risk; stronger when most of the new pages have very little text.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Many near-empty sitemap pagesA large share of the indexable pages your sitemap lists, among every page the production crawl reads, has very little main text, or many of them are nearly the same text (one empty template, locked previews).Warningwatch/thin-share
Why it matters
Many indexable pages with little in them is how Google's scaled content policy and Bing's guidelines describe low-value sites, and a site can be demoted as a whole for them.
How to fix
Keep only pages with real substance in the sitemap and index. Noindex locked previews or show enough of them, and merge or remove near-empty pages instead of adding more.
  • No Autopilot fix. Removing or noindexing pages is the owner's call; Autopilot never adds filler text.
Severity
Shown as a penalty risk lead: word counts alone are never proof, since tools and archives are short by nature.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Families of near-identical pagesThree or more pages in one language that share most of their text and differ only by a word or two in the title or address, such as a city or country, across every page the production crawl reads. Compared from text fingerprints, never the text itself.Warningwatch/templated-family
Why it matters
Pages made from one template for many places or keywords, with little else that differs, are what Google's doorway policy describes.
How to fix
Merge them into one strong page and 301 the rest, or give each page content of its own that a visitor from that search would need.
  • No Autopilot fix. Merging or removing pages is the owner's call; Autopilot never generates text to make pages look different.
Severity
Shown as penalty risk; stronger for large families that share most of their text.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Google crawling and indexing fewer pagesFrom Search Console's URL Inspection data: most inspected sitemap pages are discovered but not crawled, Google has not crawled any inspected page for weeks while new pages wait, or the indexed count is falling.Warningwatch/crawl-stall
Why it matters
A stall is a symptom, not a verdict. It often follows a site-wide quality reassessment after a spam or core update, but crawl budget, server errors and weak internal links cause it too.
How to fix
Check Search Console's Page indexing and Crawl stats reports, clean up any other penalty risk findings first, and stop adding pages until indexing recovers.
  • No Autopilot fix.
Severity
Shown as a low penalty risk lead; needs Search Console.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Key pages without a share imageThe home page or another key page has no og:image, or declares og:image:width/og:image:height below 1200×630.Infowatch/og-image-missing
Why it matters
Link previews in chat apps and social networks, and AI answers that show page images and product thumbnails, read og:image. Without a large image they show a small one, pick one themselves or show none.
How to fix
Give each key page an image of at least 1200×630 (1.91:1), or a site-wide default, and declare its width and height.
  • Autopilot only points og:image at an image that already exists in the repository; it never creates one. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans

Agentic browsing

llms.txt that fails the format check/llms.txt exists but has no # heading, fewer than 50 characters or no Markdown links.Warningagent/llms-txt-malformed
Why it matters
Lighthouse's Agentic Browsing category fails a malformed llms.txt and marks a missing one not applicable. Agents that build with your product read it; Google Search does not use it, so it is plumbing for agents, not a ranking or citation factor.
How to fix
Use a # Site name heading, a > one-line summary and lists of - [Page title](URL): what the page answers, linking only public pages.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
llms.txt that returns a web page/llms.txt answers with HTML: usually the home page, a theme's not-found page or a redirect to one.Warningagent/llms-txt-html
Why it matters
Agents that read it get a web page instead of a summary, and Lighthouse fails the file. It often comes from a plugin or a catch-all route.
How to fix
Serve a real Markdown text file at /llms.txt, or remove the route so it returns 404.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Agent catalog (ard.json) that doesn't follow the format/.well-known/ard.json exists but isn't JSON, has no entries array, or an entry lacks an urn:air: identifier, a displayName, a type or exactly one of url and data.Warningagent/ard-invalid
Why it matters
Agents skip catalog entries they can't identify or fetch, and Lighthouse audits the file. ARD is a proposal (v0.91), so the format can still change.
How to fix
Fix the entries the finding names, checking the current ARD proposal first.
  • Autopilot only edits an ard.json your site already publishes as a static file, following the current ARD proposal and keeping each entry's meaning. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Agent catalog only at the old pathThe site serves /.well-known/ai-catalog.json and not /.well-known/ard.json.Infoagent/ard-legacy-path
Why it matters
The current proposal and Lighthouse look for ard.json; agents may never find a catalog at the older path.
How to fix
Serve the same catalog at /.well-known/ard.json too.
  • Autopilot copies your existing ai-catalog.json to ard.json unchanged, when it is a static file in your repository. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
ard.json entries with too few or too many example queriesAn entry in /.well-known/ard.json lists fewer than 2 or more than 5 representativeQueries.Infoagent/ard-queries
Why it matters
Example queries tell agents when an entry is the right tool. The ARD proposal asks for 2 to 5.
How to fix
Give each entry 2 to 5 short questions it answers.
  • Autopilot only edits an ard.json your site already publishes as a static file, following the current ARD proposal and keeping each entry's meaning. It always waits for your review.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Live pages without a languageProduction pages render <html> without a lang attribute.Warningagent/html-lang-missing
Why it matters
Search engines, screen readers and AI agents use the declared language to read and quote the page correctly; Lighthouse checks it.
How to fix
Set lang once on <html> in the root layout or template.
  • The Paradigma agent opens the fix as a pull request and verifies it live. It merges on its own only if you turn on automatic merges.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans
Forms that AI agents can't use as toolsCrawled pages have forms and none declares a WebMCP tool with toolname and tooldescription.Infoagent/webmcp-forms
Why it matters
With WebMCP, a browser agent can call your form as a tool instead of guessing at the layout. Chrome runs it as an origin trial and Lighthouse lists registered tools; Shopify turned it on for Liquid stores in August 2026.
How to fix
Add toolname and tooldescription to the forms agents should use (search, booking, contact). It is optional while WebMCP is in trial.
  • No Autopilot fix. WebMCP is in an origin trial and still changing, so Paradigma suggests it instead of planning a fix.
Runs
Your monitoring schedule: every day on Pro and Agency, weekly on Free
Shows up in
The dashboard, and Slack on paid plans

Only real blockers fail the check. Everything else is a heads-up you can act on, or mute.

Blocker fails the GitHub check: the change would hide pages from Google or AI answers. Warning and info are reported without blocking the merge. Any rule can be muted for a repository from the check run, and Paradigma never pushes to your default branch.

New to the terms? The glossary explains them in a line each, and the Next.js SEO guide shows where they come from in an App Router app.

See what AI says about you. Free to start. No credit card.

Claude CodeCodexCursor