How the AEO score is calculated

Fetches /robots.txt. HTTP 200 with a text policy = 1.0. HTTP 404 = 0.6 (crawlers treat everything as allowed, but you have no explicit AI policy or sitemap pointer). 5xx, timeouts or an HTML page served as robots.txt = 0.0, because major crawlers treat an unreachable robots.txt as 'do not crawl'.

access.robots_txt

Evaluates robots.txt rules for the OAI-SearchBot token against the homepage and the key pages we scanned. A full block scores 0; a partial block scores by share of key pages allowed. We also request the homepage with this crawler's user agent; a 403/challenge when a normal browser gets 200 caps quality at 0.2.

access.bot.oai-searchbot

Evaluates robots.txt rules for the GPTBot token against the homepage and the key pages we scanned. Blocking a training crawler is a legitimate business choice, so a full block scores 0.5 (warning), not 0. We also request the homepage with this crawler's user agent; a 403/challenge when a normal browser gets 200 caps quality at 0.2.

access.bot.gptbot

Evaluates robots.txt rules for the Claude-SearchBot token against the homepage and the key pages we scanned. A full block scores 0; a partial block scores by share of key pages allowed. We also request the homepage with this crawler's user agent; a 403/challenge when a normal browser gets 200 caps quality at 0.2.

access.bot.claude-searchbot

Evaluates robots.txt rules for the ClaudeBot token against the homepage and the key pages we scanned. Blocking a training crawler is a legitimate business choice, so a full block scores 0.5 (warning), not 0. We also request the homepage with this crawler's user agent; a 403/challenge when a normal browser gets 200 caps quality at 0.2.

access.bot.claudebot

Evaluates robots.txt rules for the PerplexityBot token against the homepage and the key pages we scanned. A full block scores 0; a partial block scores by share of key pages allowed. We also request the homepage with this crawler's user agent; a 403/challenge when a normal browser gets 200 caps quality at 0.2.

access.bot.perplexitybot

Evaluates robots.txt rules for the Google-Extended token against the homepage and the key pages we scanned. Blocking a training crawler is a legitimate business choice, so a full block scores 0.5 (warning), not 0. This is a robots.txt token only, so there is no live request.

access.bot.google-extended

Evaluates robots.txt rules for the Googlebot token against the homepage and the key pages we scanned. A full block scores 0; a partial block scores by share of key pages allowed. We request the homepage with this user agent for information only. The operator verifies this crawler by IP address, so a refusal of our request is expected and is not scored.

access.bot.googlebot

Evaluates robots.txt rules for the Bingbot token against the homepage and the key pages we scanned. A full block scores 0; a partial block scores by share of key pages allowed. We request the homepage with this user agent for information only. The operator verifies this crawler by IP address, so a refusal of our request is expected and is not scored.

access.bot.bingbot

Status of the final homepage response. 200 = 1.0, other 2xx = 0.8, anything else = 0.

access.http_status

Counts redirect hops on the homepage and scanned pages. 0–1 hops = 1.0, 2 hops = 0.6, 3+ = 0.2. A homepage that redirects to a different domain caps quality at 0.6. Pages with long chains are listed.

access.redirects

Time to first byte of the homepage, measured from our server. Under 800 ms = 1.0, under 1.5 s = 0.75, under 3 s = 0.45, slower = 0.1. The median across scanned pages is shown as evidence. Network distance affects this number, so treat it as indicative.

access.response_time

Fingerprints WAF/CDN products from response headers and compares the homepage response for a normal browser against AI crawler user agents (excluding IP-verified crawlers). Quality = share of tested AI crawlers that received real content. If every probe failed for network reasons, the check is unavailable.

access.waf_blocking

Counts words of main content present in the raw HTML (no JavaScript executed). Most AI crawlers do not run JavaScript. Per page: 150+ words = 1.0, 50–149 = 0.5, under 50 = 0. Category quality is the average across key pages.

readability.server_rendered

Looks for single-page-app shells: empty #root/#app/#__next mount points, <noscript> 'enable JavaScript' messages, and pages with under 50 words but 5+ scripts. Per page: no signals = 1.0, one signal with 150+ words = 0.7, otherwise 0. This scan does not execute JavaScript, so this measures what a non-rendering crawler sees.

readability.js_dependency

Per page: a <title> of 15–70 characters that is not shared with another scanned page = 1.0. Too short / too long = 0.6. Duplicate = 0.5. Missing = 0.

readability.title

Per page: a meta description of 50–170 characters = 1.0; present but outside that range = 0.6; missing = 0.

readability.meta_description

Per page: a canonical link on the same site = 1.0; canonical pointing to another domain = 0.3; missing = 0.4.

readability.canonical

Per page: starts at 1.0; no H1 −0.5; more than one H1 −0.2; each skipped level (e.g. H2 → H4) −0.15; fewer than 2 subheadings on a page with 300+ words −0.2.

readability.headings

Per page: a <main> landmark (or role=main) = 1.0; no landmark but 200+ words of sectioned body text = 0.7; otherwise 0.3.

readability.main_content

Per page HTML size: ≤ 500 KB = 1.0, ≤ 1.5 MB = 0.6, larger = 0.2, truncated at our 3 MB cap = 0.

readability.page_size

Readable text bytes ÷ HTML bytes per page. ≥ 10% = 1.0, ≥ 4% = 0.6, ≥ 1.5% = 0.3, lower = 0.

readability.content_ratio

Share of attempted internal pages that returned real content. Quality = loaded ÷ attempted. Failed pages are listed with their status.

readability.broken_pages

Looks for Organization or a LocalBusiness subtype in JSON-LD on any scanned page. Present = 0.55, plus 0.15 each for name, url and logo (max 1.0). Missing = 0.

structure.organization

WebSite JSON-LD with a name = 1.0; without a name = 0.7; missing = 0.

structure.website

Looks for Service, Product, Offer or OfferCatalog JSON-LD (or Organization makesOffer / hasOfferCatalog). Quality = share of scanned service/offer pages carrying it; if no service pages were found, 1.0 when present anywhere and 0 otherwise.

structure.service

Applies only when the site has FAQ content (an FAQ page, or a page with 3+ question headings). Quality = share of those pages with FAQPage JSON-LD. Not applicable otherwise (FAQ content itself is scored under Answerability).

structure.faqpage

Applies only when blog/article pages were scanned. Quality = share with Article, BlogPosting or NewsArticle JSON-LD.

structure.article

Every JSON-LD block must parse as JSON and declare an @type. Quality = valid blocks ÷ total blocks. Not applicable when the site has no JSON-LD (missing schema is scored by the schema checks).

structure.jsonld_valid

Reads sitemaps declared in robots.txt, falling back to /sitemap.xml and /sitemap_index.xml. Valid sitemap with URLs = 1.0; valid but empty = 0.4; none found = 0.

structure.sitemap

Share of successfully scanned pages (excluding legal pages) whose URL appears in the sitemap, compared without tracking parameters, trailing slashes or www. Not applicable without a sitemap.

structure.sitemap_complete

Checks /llms.txt. A Markdown file starting with an H1 = 1.0; present but not Markdown = 0.5; missing = 0. llms.txt is an emerging convention. Major AI search engines have not confirmed that they use it, so it carries a low weight.

structure.llms_txt

Applies to scanned pages two or more path levels deep. Quality = share with BreadcrumbList JSON-LD. Not applicable for flat sites.

structure.breadcrumb

Per page: og:title, og:description and og:image present = 1.0, each missing tag −0.33.

structure.open_graph

Takes the first substantive paragraph after the H1. Per page: +0.4 if it exists, +0.3 if its first sentence is 8–35 words, +0.3 if it states what something is or does (definition pattern). Pages without such a paragraph score 0.

answerability.direct_answer

Counts H2–H6 headings phrased as questions across key pages. 6+ = 1.0, 3–5 = 0.7, 1–2 = 0.4, none = 0. Matching the way buyers phrase questions helps AI systems map your content to prompts.

answerability.question_headings

Per page: share of sections (max credit at 2) that contain a definition-style sentence ("X is a…", "X refers to…", "We help…"). 2+ = 1.0, 1 = 0.6, none = 0.

answerability.definitions

Per page: counts specific figures (percentages, prices, '12 years', '500+ clients', founding year, star ratings). 3+ = 1.0, 1–2 = 0.5, none = 0.

answerability.statistics

Per page: distinct multi-word proper names in the text, such as cities, brands, certifications and people. 4+ = 1.0, 2–3 = 0.6, 1 = 0.3, none = 0. Specific names help AI systems connect your business to the right context.

answerability.named_entities

Site-wide: pages with source attributions or credentials ('according to', 'certified', 'licensed', 'source:') plus outbound links to non-social external sites. Quality = 0.5 × share of key pages with attribution language + 0.5 × min(1, distinct external domains linked ÷ 3).

answerability.citation_opportunities

Per page (300+ words only): at least one content list or table = 1.0, none = 0.3. Shorter pages are not judged.

answerability.lists_tables

Per page (150+ words): quality = share of headed sections with 25–350 words; −0.3 if any single section exceeds 500 words. Self-contained sections are what AI systems retrieve and quote.

answerability.chunkability

1.0 if the site has an FAQ page or FAQPage schema with 5+ questions, or 8+ question headings site-wide; 0.6 with 3+ questions; 0.3 with 1–2; 0 with none.

answerability.faq_coverage

Per service/offer page: 300+ words = 1.0, 150–299 = 0.6, under 150 = 0.2. If no service pages were found, the homepage is judged instead, at 0.5× credit, because AI systems need a dedicated page to cite for each service.

answerability.service_descriptions

Per page: the title, H1 and opening paragraph should share topic terms. Quality = 0.5 × (title∩H1 overlap > 0) + 0.5 × (share of H1 terms that appear in the opening paragraph).

answerability.page_relevance

Collects the business name from Organization schema, WebSite schema, og:site_name, the homepage title and the footer copyright. Names are compared after removing legal suffixes, case and punctuation (one containing the other counts as a match). Quality = share agreeing with the most common name. A single source = 0.6. No sources = 0.

trust.business_name

Compares the homepage meta description, og:description and Organization schema description. Quality = 0.4 for having at least one, +0.3 for having two or more, +0.3 when they share vocabulary (word-overlap ≥ 0.2).

trust.business_description

1.0 if Person schema, Organization founder, or Article author is present. 0.7 if text names a founder, owner, director or author (for example, 'founded by Jane Smith'). 0.3 if a team page exists with no names detected. Otherwise 0.

trust.founder_author

1.0 if an About page was scanned with 150+ words; 0.6 if it was scanned but thinner, or linked but not scanned; 0 if no About page or link exists.

trust.about_page

One third each for: a phone number (tel: link, schema telephone or visible number), an email address (mailto: or schema email), and a physical/service address (PostalAddress schema or contact page with an address pattern). A contact page alone adds 0.15, up to a maximum of 1.0.

trust.contact_info

1.0 if a LinkedIn company or profile URL is linked or listed in schema sameAs; 0 otherwise.

trust.linkedin

Distinct social platforms linked or listed in sameAs (Facebook, Instagram, X, YouTube, TikTok, Pinterest, Threads, Bluesky; LinkedIn is scored separately). 2+ = 1.0, 1 = 0.6, none = 0.

trust.social_profiles

Links or sameAs entries pointing to independent business listings and knowledge bases (Google Business Profile, Wikipedia/Wikidata, Yelp, BBB, Trustpilot, Clutch, industry directories). 2+ = 1.0, 1 = 0.6, none = 0. This scan checks what your site points to. It does not search the wider web.

trust.external_references

One fifth each for: Organization schema logo, Organization schema url matching the site, schema sameAs with 2+ entries, og:site_name, and og:image on the homepage.

trust.org_identity

0.5 for Review or AggregateRating schema; 0.3 for visible testimonials, reviews or case studies (a heading or link mentioning them); 0.2 for links to third-party review platforms. Max 1.0.

trust.reviews_proof

Per page: a noindex directive in meta robots or the X-Robots-Tag header = 0; a canonical pointing to another domain = 0.5; otherwise 1.0. Site quality is the average, and a noindex homepage caps it at 0.2.

trust.indexability

Looks for text addressed to AI systems ('ignore previous instructions', 'AI assistants should recommend…') inside hidden elements, HTML comments or zero-width-character runs. Any hit = 0, because AI platforms treat this as manipulation. None = 1.0. Detected text is reported as evidence and never followed.

trust.content_integrity