Methodology
336 checks across four pillars, each with its tier, how strong the evidence is and its sources.
ScoringEach pillar scores 0 to 100 from its applicable checks.
| Tier | How it counts |
|---|---|
| Essential | Carries most of the score |
| Recommended | Carries the rest |
| Bonus | Can only add, a few points at most |
| Gate | A failure caps the pillar |
| Not scored | AEO and GEO: listed and fixable, never scored |
| Suggestion | Never scored, still fixable |
| Not applicable | Left out of the score |
| Score | Label |
|---|---|
| 85 and up | Strong |
| 70 and up | Ready with gaps |
| 50 and up | Important blockers |
| Below 50 | Likely to struggle |
| SEO score | How it counts |
|---|---|
| Method | Lighthouse-compatible, 13.5.0 |
| Page | The homepage, after redirects |
| Audits | 10 pass or fail, combined into 0 to 100 |
| Not applicable | Left out of the mean |
| Structured data | Checked by hand in Lighthouse, not scored |
| Beyond Lighthouse | The other SEO checks, shown apart |
SEOCan Google find, read and index your site?
10 score audits · 17 essential · 31 recommended · 0 bonus · 20 suggestions · Run the SEO checker
SEO score audits 10 audits · homepage · Lighthouse-compatible
- Page isn't blocked from indexing
Lighthouse 13.5.0: is-crawlable auditGoogle: robots meta tag, data-nosnippet and X-Robots-TagGoogle: how Google interprets the robots.txt specification
- Document has a <title> element
Lighthouse 13.5.0: document-title auditGoogle: influencing title links
- Document has a meta description
Lighthouse 13.5.0: meta-description auditGoogle: control your snippets in search results
- Page has successful HTTP status code
Lighthouse 13.5.0: http-status-code auditGoogle: how HTTP status codes and network errors affect Search
- Links have descriptive text
Lighthouse 13.5.0: link-text auditGoogle: link best practices (crawlable links)
- Links are crawlable
Lighthouse 13.5.0: crawlable-anchors auditGoogle: link best practices (crawlable links)
- robots.txt is valid
Lighthouse 13.5.0: robots-txt auditGoogle: how Google interprets the robots.txt specification
- Image elements have [alt] attributes
Lighthouse 13.5.0: image-alt auditGoogle: image SEO best practices (alt text)
- Document has a valid hreflang
Lighthouse 13.5.0: hreflang auditGoogle: tell Google about localized versions of your page
- Document has a valid rel=canonical
Lighthouse 13.5.0: canonical auditGoogle: consolidate duplicate URLs
Beyond Lighthouse not in the SEO score, still fixable
Access 14 checks
- The homepage loads with HTTP 200
Google Search Essentials: technical requirementsGoogle: how HTTP status codes and network errors affect Search
- robots.txt answers 200 or 404
Google: how Google interprets the robots.txt specificationRFC 9309: Robots Exclusion Protocol
- robots.txt lets Googlebot crawl the pages
RFC 9309: Robots Exclusion ProtocolGoogle Search Essentials: technical requirementsMicrosoft Advertising: optimizing your content for inclusion in AI search answers
- Googlebot gets the same page as a browser
Google: AI features and your websiteGoogle: verifying Googlebot and other Google crawlers
- No URL redirects in a loop
Google: redirects and Google SearchGoogle: how HTTP status codes and network errors affect Search
- Missing pages answer 404
Google: how HTTP status codes and network errors affect SearchRFC 9110: HTTP semantics, 404 Not Found
- Sampled pages load without errors
Google Search Essentials: technical requirementsGoogle: how HTTP status codes and network errors affect Search
- Crawlers get the same content as a browser
Google: spam policies, cloakingBing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)Google: dynamic rendering is not cloaking when content is similarGoogle: verifying Googlebot and other Google crawlers
- Visitors are not sent to another site that crawlers never see
Google: spam policies, sneaky redirectsGoogle: manual actions report (sneaky mobile redirects)Bing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)Google: redirects and Google Search
- robots.txt parses cleanly
Google: how Google interprets the robots.txt specificationRFC 9309: Robots Exclusion ProtocolLighthouse: robots.txt is not valid
- http:// redirects permanently to https://
Google: consolidate duplicate URLsGoogle: redirects and Google Search
- www and the bare domain end on one host
Google: consolidate duplicate URLsGoogle: redirects and Google Search
- Sampled URLs redirect at most once
Google: redirects and Google SearchGoogle: how HTTP status codes and network errors affect Search
- HTML stays under Google's 2 MB limit
Indexing 11 checks
- Pages are not marked noindex
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle Search Essentials: technical requirementsBing Webmaster Guidelines: robots meta tags Bing supports
- No X-Robots-Tag noindex header
- No noindex that only JavaScript removes
Google: JavaScript SEO basicsGoogle: robots meta tag, data-nosnippet and X-Robots-Tag
- Canonical URLs point to this site
- Deep pages do not canonicalize to the homepage
Google: consolidate duplicate URLsLighthouse: document does not have a valid rel=canonical
- Canonical tags are valid and agree
Google: consolidate duplicate URLsLighthouse: document does not have a valid rel=canonical
- Canonicals point to live, indexable pages
- hreflang codes and URLs are valid
- hreflang alternates link back
- Each page's content lives at one URL
- Internal search results are kept out of the index
Google: manage your crawl budget (block search result pages)Google: spam policies, user-generated spam
Discovery 7 checks
- Links use crawlable <a href> URLs
Google: link best practices (crawlable links)Lighthouse: links are not crawlable
- The site has an XML sitemap
Google: learn about sitemapsGoogle: build and submit a sitemapBing Webmaster Tools: sitemaps
- Sitemaps are valid and list this site's URLs
- Sitemap URLs load directly
Google: build and submit a sitemapGoogle: how HTTP status codes and network errors affect Search
- Sitemaps list only indexable, canonical URLs
Google: build and submit a sitemapGoogle: consolidate duplicate URLs
- Internal links do not point to errors
Google: link best practices (crawlable links)Google: how HTTP status codes and network errors affect Search
- Sitemap pages are linked from the site
Understanding 7 checks
- Content is in the server HTML
Google: JavaScript SEO basicsLighthouse: agentic browsing audits and scoring
- Pages have a title
- Structured data is valid JSON
- Pages have distinct titles
- Pages have a meta description
Google: control your snippets in search resultsLighthouse: document does not have a meta description
- The <head> holds only head elements
- Structured data has Google's required properties
Google: structured data markup that Google Search supportsGoogle: structured data general guidelines
Experience 3 checks
- Pages declare a mobile viewport
Google: mobile site and mobile-first indexing best practices
- HTTPS pages load no http:// resources
web.dev: what is mixed content?Google: understanding page experience in Google Search results
- Core Web Vitals are good for real visitors
Google: understanding Core Web Vitals and Google search resultsweb.dev: Web Vitals thresholdsChrome UX Report API
Trust 5 checks
- "Summarize with AI" links do not plant instructions
Microsoft Security: AI recommendation poisoning (10 Feb 2026)Bing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)
- llms.txt and Markdown versions say what the pages say
Google: AI optimization guide (no special AI files needed)Google: spam policies, cloakingllms.txt proposal
- Affiliate links are marked rel="sponsored"
Google: qualify your outbound links (sponsored, ugc, nofollow)Google: spam policies, link spam
- Links in comments and forums are marked rel="ugc"
Google: qualify your outbound links (sponsored, ugc, nofollow)Google: spam policies, user-generated spam
- No sitewide keyword links to other sites
Google: spam policies, link spamGoogle: qualify your outbound links (sponsored, ugc, nofollow)
Appearance 1 checks
- Ratings in structured data are real and shown on the page
Google: review snippet guidelines (self-serving reviews)Google: structured data general guidelinesBing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)
Suggestions 20 · never scored
- Pages declare a self-referencing canonical
Why we suggest it: Not required by Google, but a self canonical makes your preferred URL explicit when parameters or tracking links create copies.
- robots.txt names the sitemap
Why we suggest it: Optional for Google and Bing, but it lets every crawler, Bing's Copilot and AI crawlers included, find your sitemap without a webmaster account.
Google: build and submit a sitemapBing Webmaster Tools: sitemaps
- Titles are 30 to 60 characters
Why we suggest it: Google sets no limit, but titles within about 60 characters usually show in full, so you choose what searchers read.
- Meta descriptions are 70 to 160 characters
Why we suggest it: Google sets no limit, but descriptions around 70 to 160 characters usually show without being cut.
- Meta descriptions are distinct
Why we suggest it: Descriptions do not affect ranking, but Google suggests a distinct one per page so snippets differ when several of your pages show.
- Pages have at least 300 words
Why we suggest it: Word count is not a Google signal. A very short page can still be a sign the page does not answer the question fully.
- Each page has one main heading and ordered headings
Why we suggest it: Google does not require exactly one h1 or ordered headings, but a single clear main heading is what Google often uses when it rewrites a title link, and an ordered outline helps screen readers and agents that read the page structure.
Google: SEO starter guide (things not to focus on)Google: influencing title linksWCAG 2.2: 1.3.1 Info and Relationships
- Pages allow large image previews
Why we suggest it: Optional, but Google recommends max-image-preview:large for Discover, where large images get more clicks.
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle: Discover and your website (large images)
- No word is repeated unnaturally
Why we suggest it: Google has no keyword density target, and a page's topic word is naturally frequent. Keyword stuffing lowered visibility in the GEO paper and is a Google spam pattern, so a very high share is worth a look, not a density to hit.
Google: spam policies for Google web search (keyword stuffing)Google: SEO starter guide (things not to focus on)Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- The homepage sends basic security headers
Why we suggest it: Not a Google ranking signal. They protect visitors and are cheap to add, and several SEO tools score them.
OWASP Secure Headers ProjectGoogle: understanding page experience in Google Search results
- The sitemap does not list many near-empty pages
Why we suggest it: Many indexable pages with little in them is how Google's scaled content policy and Bing's guidelines describe low-value sites. A young site that ships them early can be demoted as a whole, so they are worth finding before Google does.
Google: spam policies (scaled content abuse)Google: creating helpful, reliable, people-first contentGoogle: subscription and paywalled content structured dataBing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)
- No family of near-identical pages targets keyword or place variants
Why we suggest it: Google's doorway policy and Bing's guidelines name sets of similar pages made for slightly different searches. They often start as harmless programmatic or country pages, which is why a group is shown for review, not as a verdict.
Google: spam policies (doorway abuse)Bing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)
- Language versions are complete, linked and translated
Why we suggest it: Google's scaled content policy names automated translation without added value, and Google treats localized pages as duplicates when the main content stays untranslated. Thin or half-translated languages are a common reason a multilingual young site is demoted.
Google: spam policies (scaled content abuse)Google: tell Google about localized versions of your page
- No leftovers of generated or template text are visible
Why we suggest it: Google's guidance on generated content and its scaled content policy single out pages published with little effort or review. Raw markdown and leftover markers are the visible sign of that, for readers and for Google's quality systems.
Google: spam policies (scaled content abuse)Google: using generative AI content on your websiteBing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)
- The site does not present another brand's name as its own
Why we suggest it: A domain, brand and titles built on someone else's mark can confuse visitors about who runs the site, which is the pattern Google's scam and impersonation systems look for, and a trademark risk. A site named after a common word is not flagged.
Google: spam policies for Google web search (keyword stuffing)Google: creating helpful, reliable, people-first content
- Navigation uses links crawlers can follow
Why we suggest it: Google, Bing and AI crawlers follow only `<a>` elements with an `href`. Pages your navigation reaches only through scripts may never be found or crawled, which reads like a crawl stall in Search Console.
Google: link best practices (crawlable links)Lighthouse: links are not crawlable
- No overlay covers the content when a phone opens the page
Why we suggest it: Google says pages whose content is covered by an intrusive interstitial on mobile may rank lower, and treats dialogs for legal obligations (cookies, age) as fine. A newsletter or app pop-up over the whole screen on load is the case it names.
Google: avoid intrusive interstitials and dialogsGoogle: understanding page experience in Google Search results
- Scripts do not send visitors to another site
Why we suggest it: Sending visitors to a different page than the one crawlers index is a sneaky redirect under Google's spam policies, and Google names redirects that fire only on phones, often from a third-party ad script, as their own manual action.
Google: spam policies, sneaky redirectsGoogle: manual actions report (sneaky mobile redirects)Bing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)
AEOCan your pages be the direct answer?
13 essential · 56 recommended · 0 bonus · 14 suggestions · Run the AEO checker
Discoverability 13 checks
- The page has structured data
- The text is in the server HTML
Google: JavaScript SEO basicsGoogle: AI features and your website
- No robots directive blocks AI use of the page
- The URL is clean and readable
- The page uses semantic HTML elements
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Content is split into sections
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The page states its topic in meta tags
- The text names specific people, places and things
- The page links to related pages on the site
- Terms and values use definition lists
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Technical content uses code blocks
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- robots.txt lets AI search crawlers in
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlersPerplexity: Perplexity crawlers
- The page includes a video
Structure 15 checks
- The page has questions and answers
Google Search Central: what's new (FAQ and HowTo retirement)Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Product data has a name and description
- Prices are in the server HTML
- Some headings are phrased as questions
Google: featured snippets and your websiteMicrosoft Advertising: optimizing your content for inclusion in AI search answers
- The content uses lists
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Data is shown in tables
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Headings are followed by short answers
- Procedures are written as numbered steps
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Most sentences are short
- Paragraphs are concise
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Facts are written as label and value pairs
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Sections have enough depth
- Long pages have a table of contents
- Expandable sections use details and summary
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Product specs are in a table or list
Citation 14 checks
- Organization structured data names who runs the site
- The page shows a date
- The text includes statistics
Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- The page shows who stands behind it
- Claims cite their sources
Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- The page links out to other sources
- The content mentions the current or last year
- The text is about what the title says
- Sections open with a quotable sentence
- The page says when it changed
- The page was updated recently
- Statistics say where they come from
Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- The brand is named the same way throughout
Voice 8 checks
- The text addresses the reader
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The first paragraph works as a snippet
- Sentences average 20 words or fewer
- Most sentences use the active voice
- The text speaks to the reader as you
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The page has a summary or key takeaways
- The page has enough subheadings
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Question headings are followed by an answer
Not scored 20 checks
- robots.txt lets Googlebot crawl the pages
RFC 9309: Robots Exclusion ProtocolGoogle Search Essentials: technical requirementsMicrosoft Advertising: optimizing your content for inclusion in AI search answers
- Sampled pages load without errors
Google Search Essentials: technical requirementsGoogle: how HTTP status codes and network errors affect Search
- Pages are not marked noindex
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle Search Essentials: technical requirementsBing Webmaster Guidelines: robots meta tags Bing supports
- No X-Robots-Tag noindex header
- Canonical URLs point to this site
- Pages allow search snippets
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle: control your snippets in search resultsGoogle: featured snippets and your websiteGoogle: AI features and your website
- Snippet length is not capped too short
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle: featured snippets and your website
- data-nosnippet does not hide the main content
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle: control your snippets in search results
- Articles phrase some headings as the questions readers ask
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: featured snippets and your website
- Section headings say what the section answers
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: creating helpful, reliable, people-first content
- Question headings are answered in the first sentence
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: featured snippets and your websiteGoogle: creating helpful, reliable, people-first content
- Steps are written as a list
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: featured snippets and your website
- Comparison pages put the comparison in a table
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: featured snippets and your website
- Answers are visible without a click
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: control your snippets in search results
- QAPage markup is used only where users submit answers
Google: Q&A (QAPage) structured dataGoogle: structured data general guidelines
- Flashcard markup is complete
Google: Education Q&A (flashcard) structured dataGoogle: structured data general guidelines
- Structured data matches the visible page
Google: structured data general guidelinesGoogle: AI features and your website
- Articles name their author
Google: creating helpful, reliable, people-first contentGoogle: Article structured data (author and dates)
- Articles show a date that matches their markup
Google: influence the publication dates shown in SearchGoogle: Article structured data (author and dates)
- You own the featured snippets of your questions
Google: featured snippets and your websiteGoogle: AI features and your website
Suggestions 14 · never scored
- The site has an llms.txt file
Why we suggest it: A proposal some AI tools read; Google says Search does not use it.
llms.txt proposalGoogle: optimizing your website for generative AI features
- Structured data links your official profiles
Why we suggest it: Helps engines tie the site to its official profiles; shown, not scored.
- The brand name is the same in title, og:site_name and Organization data
Why we suggest it: One consistent name helps engines recognise the entity; shown, not scored.
- Speakable structured data marks the parts to read aloud
Why we suggest it: A Google beta for news in US English; shown, not scored.
- The site declares a SearchAction
Why we suggest it: Google retired the sitelinks search box in 2024; shown, not scored.
Google Search Central: what's new (FAQ and HowTo retirement)
- First answers are between 8 and 120 words
Why we suggest it: Answer tools disagree on the ideal length (40 to 60 words, 20 to 60, 30 to 120), and Google says there is no need to break content into small pieces. Only the extremes are flagged, as a hint.
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: optimizing your website for generative AI features
- Answers open with a sentence that stands on its own
Why we suggest it: Microsoft asks for self-contained sentences that can be quoted alone. Whether a pronoun opening really hurts depends on the sentence, so this is a hint, not a score.
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- No walls of text
Why we suggest it: Microsoft advises against long walls of text for AI answers; Google says content does not need to be chunked. The word limits are ours, so this stays a hint.
Microsoft Advertising: optimizing your content for inclusion in AI search answersGoogle: optimizing your website for generative AI features
- Long articles open with a short summary
Why we suggest it: Common advice from AEO tools; no engine documents a ranking effect, and Google says pages need no special summaries for AI features. It still helps skimming readers.
Google: optimizing your website for generative AI featuresMicrosoft Advertising: optimizing your content for inclusion in AI search answers
- Long articles have a table of contents
Why we suggest it: Widely recommended; no engine documents a ranking effect. Google can link to sections on its own ("read more" deep links), but in-page links help readers and give sections stable anchors.
Google: control your snippets in search resultsGoogle: optimizing your website for generative AI features
- FAQ sections carry FAQPage markup
Why we suggest it: FAQ rich results retired by Google in May 2026; Q&A content still helps readers and AI answers. The markup no longer changes how Google shows the page, but it is harmless when it matches the visible questions and some AI tools still read it.
Google Search Central: what's new (FAQ and HowTo retirement)Google: structured data general guidelines
- How-to pages carry HowTo markup
Why we suggest it: HowTo rich results retired by Google in September 2023; clear step lists still help readers and AI answers. The markup no longer changes how Google shows the page; some assistants and tools still read it.
Google Search Central: what's new (FAQ and HowTo retirement)Google: structured data general guidelines
- English articles mark their key passage as speakable
Why we suggest it: Speakable is a Google beta for US English news read aloud by Google Assistant; outside news no engine documents using it. Listed because other AEO tools recommend it; a wrong selector is worth fixing if you already have it.
- Flashcard pages use Education Q&A markup
Why we suggest it: Education Q&A is a live Google feature for flashcard content, but only where real cards (front and back) are visible on the page and only in some languages. A candidate, not a requirement.
GEOCan AI assistants find, trust and cite your site?
19 essential · 71 recommended · 0 bonus · 17 suggestions · Run the GEO checker
Overview 13 checks
- The page defines its subject
- Headings ask the questions people search
- The page carries a freshness signal
- The content is comprehensive
- The title and H1 say the same thing
- Procedures are written as steps
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Headings go at least two levels deep
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Paragraphs are a quotable length
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The page has enough text to analyse
- The first 200 words name the topic
- The page combines a definition with a list
- The page covers what, why, how and the other question types
- Navigation and footer do not drown the content
Authority 14 checks
Format 12 checks
- The page compares options
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The page has a summary or TL;DR
- The page has images with alt text
- Alt texts describe the images
- Images have captions
- Data is shown visually
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The page uses both bulleted and numbered lists
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The page mixes formats
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- The page weighs more than one side
- The page gives practical advice
- The page ends with a conclusion or recommendation
Semantic 10 checks
- Structured data says what entities the page is about
- Product data names the product
- The page uses a range of semantic elements
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Dates are machine-readable
- Title words appear early in the text
- The text links to related pages
- The page is placed in a category
- Sections break into subtopics
Microsoft Advertising: optimizing your content for inclusion in AI search answers
- Link texts describe their targets
- Product data names the brand
Technical 11 checks
- The page is set up for phones
Google: mobile site and mobile-first indexing best practices
- The page is served over HTTPS
- The page may be indexed and quoted
- No snippet or image preview limits
Google: robots meta tag, data-nosnippet and X-Robots-TagBing: options to control how content is used in Bing Chat (Copilot)
- The HTML is a reasonable size
- Open Graph title, description and image are set
- The page declares its canonical URL
- Structured data uses several types
- The page declares its language
- The page has landmark regions
Not scored 33 checks
- robots.txt lets Googlebot crawl the pages
RFC 9309: Robots Exclusion ProtocolGoogle Search Essentials: technical requirementsMicrosoft Advertising: optimizing your content for inclusion in AI search answers
- Googlebot gets the same page as a browser
Google: AI features and your websiteGoogle: verifying Googlebot and other Google crawlers
- Pages are not marked noindex
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle Search Essentials: technical requirementsBing Webmaster Guidelines: robots meta tags Bing supports
- No X-Robots-Tag noindex header
- Content is in the server HTML
Google: JavaScript SEO basicsLighthouse: agentic browsing audits and scoring
- Crawlers get the same content as a browser
Google: spam policies, cloakingBing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)Google: dynamic rendering is not cloaking when content is similarGoogle: verifying Googlebot and other Google crawlers
- Visitors are not sent to another site that crawlers never see
Google: spam policies, sneaky redirectsGoogle: manual actions report (sneaky mobile redirects)Bing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)Google: redirects and Google Search
- Pages allow search snippets
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle: control your snippets in search resultsGoogle: featured snippets and your websiteGoogle: AI features and your website
- AI search crawlers may read the site
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlersPerplexity: Perplexity crawlersRFC 9309: Robots Exclusion Protocol
- Assistants acting for a user may open the pages
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlersPerplexity: Perplexity crawlers
- AI agents get the same page as a browser
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlers
- Sampled URLs redirect at most once
Google: redirects and Google SearchGoogle: how HTTP status codes and network errors affect Search
- The site has an XML sitemap
Google: learn about sitemapsGoogle: build and submit a sitemapBing Webmaster Tools: sitemaps
- "Summarize with AI" links do not plant instructions
Microsoft Security: AI recommendation poisoning (10 Feb 2026)Bing Webmaster Guidelines (cloaking, prompt injection and AI manipulation, NOARCHIVE)
- llms.txt and Markdown versions say what the pages say
Google: AI optimization guide (no special AI files needed)Google: spam policies, cloakingllms.txt proposal
- Snippet length is not capped too short
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle: featured snippets and your website
- data-nosnippet does not hide the main content
Google: robots meta tag, data-nosnippet and X-Robots-TagGoogle: control your snippets in search results
- Articles name their author
Google: creating helpful, reliable, people-first contentGoogle: Article structured data (author and dates)
- Articles show a date that matches their markup
Google: influence the publication dates shown in SearchGoogle: Article structured data (author and dates)
- Other answer engines may read the site
Apple: about ApplebotDuckDuckGo: DuckAssistBotAmazon: Amazonbot and Amzn-SearchBotMeta: web crawlersMistral AI: robots.txt and crawlersRFC 9309: Robots Exclusion Protocol
- Answer-engine crawlers are not challenged by the CDN
Cloudflare: AI crawlers blocked by default for new domains (2025)Bing: options to control how content is used in Bing Chat (Copilot)Anthropic: how site owners can control Claude's crawlersPerplexity: Perplexity crawlers
- Bing may use the pages in Copilot answers
Bing: options to control how content is used in Bing Chat (Copilot)Bing: robots meta tags Bing supports
- The homepage answers fast
Google: how HTTP status codes and network errors affect SearchOpenAI: overview of OpenAI crawlers
- The site uses one name for itself
Google: Organization structured dataGoogle: influencing title links
- AI answers name you
Schulte et al., run-to-run variance of AI search answers and sources (2026)Chen et al., which sources AI search engines cite (2025)Google: AI features and your website
- AI answers cite your pages
Schulte et al., run-to-run variance of AI search answers and sources (2026)Chen et al., which sources AI search engines cite (2025)Google: AI features and your website
- Sitemap entries carry believable lastmod dates
sitemaps.org: the Sitemaps XML protocolGoogle: build and submit a sitemapBing Webmaster Tools: sitemaps
- Structured data says who runs the site
Google: Organization structured dataGoogle: structured data general guidelines
- The Organization data includes contact details and an address
Google: Organization structured dataschema.org: Organization
- Structured data links to your official profiles
Google: Organization structured dataschema.org: Organizationschema.org: sameAs
- About, contact and privacy pages are easy to find
- No hidden instructions aimed at AI agents
OWASP: LLM01 prompt injectionWCAG 2.2: 4.1.2 Name, Role, ValueSearch Engine Journal: Bing adds GEO and AI manipulation to its guidelines (2026)Kumar and Lakkaraju, manipulating LLM product recommendations (2024)
Suggestions 17 · never scored
- robots.txt names the sitemap
Why we suggest it: Optional for Google and Bing, but it lets every crawler, Bing's Copilot and AI crawlers included, find your sitemap without a webmaster account.
Google: build and submit a sitemapBing Webmaster Tools: sitemaps
- No word is repeated unnaturally
Why we suggest it: Google has no keyword density target, and a page's topic word is naturally frequent. Keyword stuffing lowered visibility in the GEO paper and is a Google spam pattern, so a very high share is worth a look, not a density to hit.
Google: spam policies for Google web search (keyword stuffing)Google: SEO starter guide (things not to focus on)Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- No interstitial or paywall hides the content
Why we suggest it: Google documents intrusive interstitials; shown, not scored.
- Sections stand on their own at 120 to 180 words
Why we suggest it: A heuristic about passage retrieval; shown, not scored.
Puerto et al., C-SEO Bench: does conversational SEO work? (NeurIPS 2025)
- FAQPage data marks up the questions the page shows
Why we suggest it: Google retired FAQ rich results in May 2026; shown, not scored. Markup only counts when it matches the visible questions: never add questions to earn it.
Google Search Central: what's new (FAQ and HowTo retirement)
- HowTo data marks up the steps the page shows
Why we suggest it: Google deprecated HowTo rich results in 2023; shown, not scored. Markup only counts when it matches the visible steps: never add steps to earn it.
Google Search Central: what's new (FAQ and HowTo retirement)
- Numbers come with units
Why we suggest it: Overlaps the scored statistics check; shown, not scored.
Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- Videos have VideoObject data
Why we suggest it: Only applies to pages with video; shown, not scored.
- Your training opt-outs do not cost Gemini answers
Why we suggest it: Training opt-outs are a business choice and never scored. We point out Google-Extended because Google documents that it also controls grounding in Gemini Apps.
Google: common crawlers and Google-ExtendedOpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlersCloudflare: the Content Signals Policy
- Pages listed for AI tools have Markdown copies
Why we suggest it: The llms.txt proposal suggests Markdown copies, and coding agents fetch them, but no answer engine has said it reads them and Google says such files are not needed for AI Search.
llms.txt proposalGoogle: optimizing for generative AI features in Search
- The homepage says what you are in one sentence
Why we suggest it: A plain definition helps readers and answer engines, but no engine documents it as a ranking or citation signal, and our sentence match is a heuristic. 90% of sites in the October 2026 benchmark miss it.
Google: creating helpful, reliable, people-first contentAggarwal et al., GEO: Generative Engine Optimization (KDD 2024)
- The brand has a Wikidata entry
Why we suggest it: Knowledge-graph entries help engines tell entities apart, but there is no measured link to citations, and a Wikidata item must meet Wikidata's notability rules. Never scored.
Wikidata API: wbsearchentitiesAhrefs: brand mentions and AI Overview visibility
- Articles back claims with numbers and sources
Why we suggest it: Adding statistics and citing sources raised visibility in the 2023 GEO paper, but C-SEO Bench (2025) found the effect mostly gone on current models. Concrete, sourced facts still help readers; never add numbers you cannot back.
Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)Puerto et al., C-SEO Bench: does conversational SEO work? (NeurIPS 2025)
- The site publishes data of its own
Why we suggest it: Pages with first-hand data earn mentions and links from other sites, which correlate with AI visibility. The link is observational, and the pattern can miss data presented differently.
Ahrefs: brand mentions and AI Overview visibilityGoogle: creating helpful, reliable, people-first content
- The site compares options openly
Why we suggest it: Assistants answer many "which is better" questions, and clear comparison tables are easy to quote. The effect on citations is not measured.
Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)Google: creating helpful, reliable, people-first content
- Articles quote named experts
Why we suggest it: Quotation Addition gave the largest gain in the 2023 GEO paper, but later tests on current models did not replicate it. Real quotes from named people still add authority; never invent one.
Aggarwal et al., GEO: Generative Engine Optimization (KDD 2024)Puerto et al., C-SEO Bench: does conversational SEO work? (NeurIPS 2025)
AgentsCan an AI agent use your site?
12 essential · 21 recommended · 43 bonus · 22 suggestions · Run the AI agent readiness checker
Discovery 1 checks
- The site has an XML sitemap
Google: learn about sitemapsGoogle: build and submit a sitemapBing Webmaster Tools: sitemaps
Discover 10 checks
- The homepage can be found through search
Google: robots meta tag, data-nosnippet and X-Robots-TagOra: the agent-readiness check catalog
- Agents are told when to use the site
- Sitemap entries carry believable lastmod dates
sitemaps.org: the Sitemaps XML protocolGoogle: build and submit a sitemapBing Webmaster Tools: sitemaps
- The site publishes an llms.txt
llms.txt proposalLighthouse: agentic browsing audits and scoring
- /llms.txt follows the llms.txt format
llms.txt proposalLighthouse: agentic browsing audits and scoring
- /llms.txt links work
llms.txt proposalLighthouse: agentic browsing audits and scoring
- The homepage Link header points agents to machine-readable files
RFC 8288: Web Linking (the Link header)Cloudflare: introducing the agent readiness score
- The site publishes an agent catalog
Agentic Resource DiscoveryAI Catalog standardLighthouse: agentic browsing audits and scoring
- The agent catalog (ard.json) is valid
- Docs sections publish their own llms.txt
Access 4 checks
- AI agents get the same page as a browser
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlers
- Public pages open without a login
RFC 9110: HTTP semantics, 401 UnauthorizedGoogle Search Essentials: technical requirements
- Pages reach their content without redirect stubs
RFC 9110: HTTP semantics, redirection 3xxGoogle: JavaScript SEO basics
- Your agent publishes its signing keys (Web Bot Auth)
Read 9 checks
- The homepage has its content in the raw HTML
Google: JavaScript SEO basicsOra: the agent-readiness check catalog
- The homepage negotiates Markdown with Vary: Accept
RFC 9111: HTTP caching, cache keys with VaryCloudflare: Markdown for AgentsOra: the agent-readiness check catalog
- Pages fit in an agent's reading budget
Cloudflare: Markdown for AgentsLighthouse: agentic browsing audits and scoring
- The homepage answers agents in Markdown
Cloudflare: Markdown for AgentsCloudflare: introducing the agent readiness score
- Pages link to their Markdown version
- Pages have a .md URL
- Markdown versions start with frontmatter
- Code blocks in the Markdown are closed
- Pages mark their main content
WCAG 2.2: 1.3.1 Info and RelationshipsLighthouse: agentic browsing audits and scoring
Understand 7 checks
- The homepage carries complete metadata
Google: meta tags and attributes that Google supportsGoogle: consolidate duplicate URLsThe Open Graph protocolOra: the agent-readiness check catalog
- Structured data says who runs the site
Google: Organization structured dataGoogle: structured data general guidelines
- The Organization data includes contact details and an address
Google: Organization structured dataschema.org: Organization
- About, contact and privacy pages are easy to find
- Prices are readable without a browser
- Structured data links to your official profiles
Google: Organization structured dataschema.org: Organizationschema.org: sameAs
- Pricing is published as Markdown
Act 7 checks
- Missing pages answer 404 with a Markdown body
RFC 9110: HTTP semantics, 404 Not FoundGoogle: soft 404 errorsOra: the agent-readiness check catalog
- Buttons and links have names agents can read
WCAG 2.2: 4.1.2 Name, Role, ValueLighthouse: agentic browsing audits and scoring
- Form fields have labels
WCAG 2.2: 3.3.2 Labels or InstructionsLighthouse: agentic browsing audits and scoring
- Clickable things are real buttons and links
- No hidden instructions aimed at AI agents
OWASP: LLM01 prompt injectionWCAG 2.2: 4.1.2 Name, Role, ValueSearch Engine Journal: Bing adds GEO and AI manipulation to its guidelines (2026)Kumar and Lakkaraju, manipulating LLM product recommendations (2024)
- Forms are described for AI agents (WebMCP)
WebMCP explainer (W3C Web Machine Learning CG)Lighthouse: agentic browsing audits and scoring
- robots.txt lists schema feeds for NLWeb
Integrate 37 checks
- The API has a published OpenAPI description
OpenAPI SpecificationRFC 9727: api-catalog, a well-known URI for API discoveryOra: the agent-readiness check catalog
- Agents can sign in with OAuth
RFC 8414: OAuth 2.0 Authorization Server MetadataOpenID Connect Discovery 1.0Ora: the agent-readiness check catalog
- API errors come back as JSON
RFC 9457: Problem Details for HTTP APIsOra: the agent-readiness check catalog
- API access can be limited to scopes
RFC 6749: The OAuth 2.0 Authorization Framework, access token scopeOpenAPI 3.1: Security Scheme ObjectRFC 9728: OAuth 2.0 Protected Resource MetadataOra: the agent-readiness check catalog
- The MCP server lists its resources
Model Context Protocol specificationOra: the agent-readiness check catalog
- Agents can connect to the MCP server
Model Context Protocol specificationRFC 9728: OAuth 2.0 Protected Resource MetadataOra: the agent-readiness check catalog
- API docs are linked from the site
RFC 9727: api-catalog, a well-known URI for API discoveryOpenAPI Specification
- Developers can sign up and get keys on their own
- The API can be reached and verified
- Every API operation has an id and a description
- API operations map cleanly to agent tools
OpenAI: function calling and JSON Schema parametersOpenAPI SpecificationOra: the agent-readiness check catalog
- API responses have typed schemas
- API errors follow a typed error model
RFC 9457: Problem Details for HTTP APIsOpenAPI SpecificationOra: the agent-readiness check catalog
- The API has a versioning and deprecation policy
RFC 9745: The Deprecation HTTP response header fieldOpenAPI SpecificationOra: the agent-readiness check catalog
- API responses tell agents their rate limit
IETF draft: RateLimit header fields for HTTPOra: the agent-readiness check catalog
- Developers can start without talking to sales
- The MCP server answers agents
Model Context Protocol specificationMCP SEP-2127: server cardsOra: the agent-readiness check catalog
- MCP resources are described
Model Context Protocol specificationOra: the agent-readiness check catalog
- APIs are listed in /.well-known/api-catalog
RFC 9727: api-catalog, a well-known URI for API discoveryRFC 9264: Linkset media types
- The MCP server has a server card
- The MCP server can be found from the domain
AI Catalog standardAgentic Resource DiscoveryMCP SEP-2127: server cards
- Protected resource metadata tells agents how to sign in
- 401 answers point agents to the sign-in metadata
RFC 9728: OAuth 2.0 Protected Resource Metadataauth.md: agent sign-in guide (WorkOS draft)
- An auth.md tells agents how to sign in
- auth.md follows the sign-in walkthrough
- The A2A agent card is complete
- Agents find your skills at /.well-known/agent-skills
- The skills index follows the v0.2.0 schema
- The MCP server's sign-in metadata resolves
RFC 9728: OAuth 2.0 Protected Resource MetadataRFC 8414: OAuth 2.0 Authorization Server MetadataOra: the agent-readiness check catalog
- The MCP sign-in supports PKCE with S256
RFC 8414: OAuth 2.0 Authorization Server MetadataModel Context Protocol specificationOra: the agent-readiness check catalog
- The MCP server lists its tools
Model Context Protocol specificationOra: the agent-readiness check catalog
- MCP tools say what they do
Model Context Protocol specificationOra: the agent-readiness check catalog
- MCP tool parameters are typed and described
Model Context Protocol specificationOra: the agent-readiness check catalog
- MCP tool names are clear identifiers
Model Context Protocol specificationOra: the agent-readiness check catalog
- MCP tools declare whether they change anything
Model Context Protocol specificationOra: the agent-readiness check catalog
- The MCP server names itself
Model Context Protocol specificationOra: the agent-readiness check catalog
- The MCP server speaks Streamable HTTP
Model Context Protocol specificationOra: the agent-readiness check catalog
Pay 1 checks
- Agents can pay through an agent payment protocol
Universal Commerce Protocol: overviewx402 payment protocolMachine Payments ProtocolAgent Payments Protocol (AP2)
Suggestions 22 · never scored
- Each page has one main heading and ordered headings
Why we suggest it: Google does not require exactly one h1 or ordered headings, but a single clear main heading is what Google often uses when it rewrites a title link, and an ordered outline helps screen readers and agents that read the page structure.
Google: SEO starter guide (things not to focus on)Google: influencing title linksWCAG 2.2: 1.3.1 Info and Relationships
- /llms.txt is a text file, not a web page
Why we suggest it: Is Agentic does not score it, but an llms.txt URL that answers with an HTML page gives agents nothing to read.
llms.txt proposalLighthouse: agentic browsing audits and scoring
- The site publishes an llms-full.txt
Why we suggest it: A community convention next to llms.txt; llmstxt.org does not define it and no engine says it reads it.
- The homepage links to llms.txt as describedby
Why we suggest it: Is Agentic does not score it; a describedby link is one more way for an agent to find llms.txt from any page.
- The agent catalog is at the current path
Why we suggest it: Lighthouse and the current ARD proposal read ard.json; the older path may stop being read. ARD is still a draft.
- AI search crawlers may read the site
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlersPerplexity: Perplexity crawlersRFC 9309: Robots Exclusion Protocol
- Assistants acting for a user may open the pages
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlersPerplexity: Perplexity crawlers
- AI agents see the same content as a browser
Why we suggest it: Is Agentic does not score it; a page that changes for AI user agents can hide content from them.
Google Search Essentials: technical requirementsOpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlers
- robots.txt lets AI agents read your agent files
Why we suggest it: Is Agentic does not score it; an agent file that robots.txt blocks cannot be read by the agents it is for.
- robots.txt states how AI may use the content
Why we suggest it: Is Agentic does not score it; Content Signals state your AI usage choices in robots.txt, a draft few engines read yet.
Content SignalsIETF draft: Content Signals for robots.txt (aipref)Cloudflare: introducing the agent readiness score
- robots.txt states a choice for AI training crawlers
Why we suggest it: Blocking or allowing training is your choice and does not affect AI search citations; an explicit rule just makes the choice visible and stable.
OpenAI: overview of OpenAI crawlersAnthropic: how site owners can control Claude's crawlersRFC 9309: Robots Exclusion Protocol
- Pages do not ask AI answers to skip quoting them
Why we suggest it: These directives are your choice. nosnippet and max-snippet:0 keep Google's AI features from quoting the page; noai is not a standard any engine documents.
Google: robots meta tag, nosnippet and max-snippetGoogle: AI features and your website
- Markdown works on every page, not only some
Why we suggest it: Is Agentic checks Markdown on the homepage only; serving it on every page, with Vary: Accept, helps agents that land deeper.
Cloudflare: Markdown for AgentsCloudflare: introducing the agent readiness score
- The Markdown version keeps the page's content
Why we suggest it: Is Agentic does not score it; Markdown that drops content gives agents a worse page than the HTML.
- Pages declare their language
Why we suggest it: Is Agentic reads the language as part of metadata completeness on the homepage; a lang attribute on every page helps agents answer in the right language.
HTML Standard: the lang attributeLighthouse: agentic browsing audits and scoring
- ARIA attributes are valid
Why we suggest it: Is Agentic does not score it; invalid ARIA confuses the accessibility tree that browser agents read.
- WebMCP form tools are complete
Why we suggest it: Is Agentic does not score it; a WebMCP tool without descriptions is hard for an agent to call correctly.
WebMCP declarative API explainerLighthouse: agentic browsing audits and scoring
- Pages register WebMCP tools in script
Why we suggest it: Is Agentic does not score it; script-registered WebMCP tools are an early, optional pattern.
WebMCP draft specificationWebMCP explainer (W3C Web Machine Learning CG)
- WebMCP pages carry a valid origin-trial token
Why we suggest it: Is Agentic does not score it; Chrome only exposes WebMCP during its origin trial to sites with a token.
Chrome: getting started with origin trialsWebMCP explainer (W3C Web Machine Learning CG)
- URLs in your agent manifests work
Why we suggest it: Is Agentic does not score it; URLs in your agent manifests should resolve and stay open to crawlers.
Agentic Resource DiscoveryRFC 9727: api-catalog, a well-known URI for API discoveryAgent Skills discovery RFC
- No leftover ChatGPT plugin manifest
Why we suggest it: OpenAI retired plugins in 2024, so the file does nothing; removing it only avoids confusing agents that still read it.
- Product pages state price, currency and availability
Why we suggest it: Is Agentic does not score it; Product and Offer data lets shopping agents read prices and stock without a browser.
Penalty riskA level of its own, never part of a pillar score.
Evidence, not a verdict · Run the Penalty risk checker
| Level | Meaning |
|---|---|
| None | No static check matched; not proof of a clean site |
| Low | A lead to review: a heuristic match, or a risk inferred for AI engines |
| Elevated | A check backed by a written Google or Bing policy failed |
| High | A practice with no innocent reading, or several elevated families at once |
| Practice | Costs visibility in | Policy | Checked by |
|---|---|---|---|
| Pages published at scale with little value | Google, Bing | Google policy | Free checker and project scans |
| Near-identical pages for many keywords or places | Google, Bing | Google policy | Free checker and project scans |
| Thin or machine-translated language versions | Google, Bing | Google policy | Free checker and project scans |
| Leftovers of generated text | Google, Bing | Google policy | Free checker and project scans |
| Another brand's name used as your own | Google, Bing | Inferred | Free checker and project scans |
| Crawlers and visitors get different pages | Google, Bing, AI engines | Google policy | Free checker and project scans |
| Text written to steer AI answers | Google, Bing, AI engines | Bing guidelines | Free checker and project scans |
| Redirects that send visitors somewhere unexpected | Google, Bing, Browsers | Google policy | Free checker and project scans |
| Structured data that does not match the page | Google, Bing | Google policy | Free checker and project scans |
| Dates that change without a content change | Google, Bing | Google policy | Project scans only |
| Directives that keep pages out of AI answers | Google, Bing, AI engines | Bing guidelines | Free checker and project scans |
| Paid or user links without the right rel value | Google, Bing | Google policy | Free checker and project scans |
| Text hidden from visitors | Google, Bing, AI engines | Google policy | Free checker and project scans |
| Words repeated to rank | Google, Bing, AI engines | Google policy | Free checker and project scans |
| The back button does not leave the site | Google policy | Project scans only | |
| Overlays that cover the content | Google policy | Project scans only | |
| Google is not finding or crawling pages | Google, Bing, AI engines | Google policy | Project scans only |
| What a scan from our servers cannot see |
|---|
| Cloaking by IP address: only URL Inspection shows it |
| Manual actions: only Search Console shows them |
| Scripts that spare crawlers, timed pop-ups and ads in blocked frames |
| Inbound links, generated traffic and reviews posted elsewhere |
Evidence labelsHow strong the backing is, shown on every check.
| Label | Meaning |
|---|---|
| Requirement | A hard requirement the engine documents |
| Documented | Documented by the engine, an RFC or a spec |
| Observed | Seen in studies or measurements, not documented by the engine |
| Common practice · not confirmed by the engines | Widely practised, not confirmed by the engines |