Vol. 1 · Issue 1

No. 01 · Audit framework

The AEO Audit Checklist

The 50-item framework for AI search visibility.

The AEO Report Editorial Published 2026-05-14

Most brands appearing in ChatGPT, Perplexity, Gemini, and Claude didn't get there because their content is uniquely insightful. They got there because their pages are structurally legible to AI engines — fast, semantic, entity-rich, and threaded with the structured data that lets a model cite them confidently.

Most brands not appearing don't have a content problem. They have a plumbing problem.

This is the independent 50-item AEO audit framework we use when investigating why a brand is invisible in AI search. It's organized into seven pillars, from raw discoverability to long-term measurement. Every item is concrete: either you ship it or you don't. There's no soft middle.

Tick each item you can answer yes to. Your score updates live in the bar above and saves to your browser — close the tab and come back; your work persists.

The seven pillars

01 Discoverability
02 Citability
03 Authority Signals
04 Structured Data
05 Content Optimization
06 Engine-Specific Coverage
07 Monitoring & Measurement

Pillar 01

Discoverability

Can AI engines actually reach and parse your pages?

01

Public llms.txt file at /llms.txt

Publish a structured llms.txt at your root. It acts as a treasure map — telling AI engines which content matters and where canonical versions live. Sites publishing one in 2026 are appearing in citations at measurably higher rates.
02

Robots.txt explicitly allows AI crawlers

Whitelist GPTBot, ClaudeBot, Perplexity-User, Google-Extended, OAI-SearchBot, and Bingbot. Many sites still block them by default — if you're not in the index, you're not in the answer.
03

Server response time under 800ms

AI crawlers time out faster than Googlebot. Pages slower than ~800ms get partially crawled or skipped entirely. Cache aggressively, ship static where possible, and measure with Lighthouse-on-Vercel or WebPageTest.
04

Critical content renders without JavaScript

Most AI crawlers don't execute JavaScript. If your headline, body copy, or product spec only appears after a client-side render, the engine sees an empty shell. SSR or SSG everything that matters.
05

XML sitemap submitted to Google + Bing

Bing's index powers ChatGPT search. Submit a sitemap to Bing Webmaster Tools — not just GSC. List every canonical URL, include lastmod dates, and ping when content meaningfully changes.
06

Canonical URLs declared on every page

Ambiguous canonicals confuse AI engines about which version to cite. Set <link rel='canonical'> on every page, even self-referential. Eliminate trailing-slash mismatches and query-string duplicates.
07

Pages return clean 200 status

Soft 404s — pages that look like 404s but return 200 — are particularly toxic for AI engines. Run a crawl, audit status codes, and remove or redirect anything returning 200 for a missing resource.

Pillar 02

Citability

Are your pages structured so AI can quote and attribute them?

08

Each page has exactly one H1

AI engines use H1s to identify the page's primary topic. Multiple H1s create ambiguity; missing H1s create blank-slate citations. One H1, descriptive, keyword-aligned, near the top of the document.
09

H2s and H3s use question-format phrasing

AI engines match user prompts to question-shaped sub-headings. 'What is X?', 'How does X work?', 'When should you use X?' rank dramatically better as quoted answers than nominal headings.
10

Lead paragraph answers the primary question in 50–100 words

Front-load the answer. The first paragraph after the H1 should be a self-contained, attributable response to the page's core question. This is what AI engines lift verbatim into citations.
11

Statistics and quotes are clearly attributed

Numbers without sources don't get cited. Wrap stats in '<a>' tags pointing to the primary source, and use semantic blockquote elements for quoted material. AI engines prefer attributed claims to bare assertions.
12

Lists use semantic HTML, not styled divs

Use <ul>, <ol>, and <li>. Don't fake lists with <div class='list-item'>. AI engines extract structured lists as discrete citable units — and miss them entirely when they're hidden inside non-semantic markup.
13

Tables use real <table> markup

Comparison data only gets surfaced if it's in a real table. Use <thead>, <tbody>, <th>, and <td>. Caption the table. AI engines reliably extract table content as structured comparisons — but only from real HTML tables.
14

Definitions use <dl><dt><dd> blocks

Glossary terms, technical definitions, and pricing tiers all benefit from definition list markup. AI engines treat <dl> blocks as paired key/value data — perfect for direct quotation in answer contexts.
15

Author byline is visible with credentials

AI engines weight content with named, credentialed authors over anonymous content. Show the author near the top of each piece, link to a full bio page, and include relevant credentials inline.

Pillar 03

Authority Signals

Why should an AI engine trust your content over a competitor's?

16

Author bio pages with credentials and external links

Every author needs a dedicated bio page linking to other publications, social profiles, and any speaking or media appearances. AI engines verify authority by triangulating these signals across the web.
17

Comprehensive 'About' and 'Team' pages

Beyond an 'About' tagline: show company history, founding team, contact information, and physical address if applicable. AI engines use these signals to distinguish established entities from thin affiliate sites.
18

Brand profiles on Wikipedia, Crunchbase, LinkedIn, industry directories

AI engines cross-reference your brand against external knowledge graphs. A Wikipedia page, Crunchbase profile, Wikidata entry, and LinkedIn Company page give engines confidence that you're a real, citable entity.
19

Backlinks from at least 3 high-authority domains

AI engines still respect link signals as authority proxies. Earn at least three links from DR 60+ domains in your space. Citation in journalism, podcast show notes, or research papers carry the most weight.
20

Publication date visible on every article

Hide the date and you signal evergreen — but AI engines often interpret 'no date' as 'untrustworthy.' Surface the date clearly, formatted ISO-style or natural English, near the byline.
21

'Last updated' dates surfaced for evergreen content

AI engines strongly prefer recently-updated content. Show a 'Last updated' date when materially revising a post. Trickle-update evergreen pieces quarterly to maintain freshness signals.
22

Reviewed-by or fact-checked-by attribution where applicable

Medical, financial, legal, and technical content gets a citability bump when explicitly fact-checked by a credentialed reviewer. Add a 'Reviewed by [Name, credentials]' line to high-stakes content.

Pillar 04

Structured Data

Have you handed engines the metadata they need to cite you cleanly?

23

Organization schema on homepage and About page

Provide JSON-LD Organization schema declaring your legal name, founding date, founders, contact info, sameAs links to social and Wikipedia. This is the foundation for every entity-based AI citation.
24

Article schema on every editorial post

Wrap each article in Article (or NewsArticle/BlogPosting) schema. Include datePublished, dateModified, headline, author, publisher, and image. This is the single highest-ROI schema for AI citation eligibility.
25

Author schema linked to Article schema

Each author's Person schema should be referenced from every Article they wrote. Include name, jobTitle, sameAs links to LinkedIn/Twitter, and worksFor pointing to the publisher's Organization schema.
26

FAQPage schema on pages with question-format content

If your page contains question-and-answer pairs, mark them with FAQPage schema. AI engines lift these directly into answers. Don't fake it — Google penalizes irrelevant FAQ markup, and so do AI engines.
27

BreadcrumbList schema on category and sub-pages

Breadcrumb schema helps engines understand site hierarchy and improves citation context. Implement it on every non-homepage. It's also one of the few schemas Bing actively rewards.
28

Product or Service schema where applicable

Selling a product or service? Use Product schema with offers, pricing, ratings, and availability. AI engines lean heavily on Product schema for commerce-intent queries. Without it, you're invisible.
29

All schema validates without errors

Run every page through Google's Rich Results Test and Schema.org's validator. Invalid schema is often ignored entirely — and worse, can flag your content as low quality. Fix every warning, not just errors.

Pillar 05

Content Optimization

Is the content itself optimized for how AI engines ingest information?

30

Each page targets a single primary intent

AI engines reward focused pages. A page trying to be the definitive guide to 'AEO and SEO and content marketing and analytics' will rank for none. Pick one intent per URL. Internal-link adjacent topics.
31

Substantive length — 800+ for posts, 1,500+ for pillars

AI engines reward depth, not word count for its own sake. Thin pages (sub-300 words) struggle to surface; pillar-class pages (1,500+) get cited as canonical references. Hit the depth the topic warrants.
32

Entity-rich writing — name people, companies, products explicitly

AI engines parse content for named entities. 'A major SEO tool' is invisible; 'Ahrefs' is citable. Be explicit about who, what, and where — even when it feels redundant for human readers.
33

Internal links connect related entities and topics

Build a real internal link graph. Every mention of a tool, person, or concept should link to its dedicated page when one exists. AI engines use these connections to model topical authority.
34

Reading level appropriate to the audience

Practitioner-level content reads at grade 10–12; executive content reads at grade 8–10. AI engines lift quotable lines — if every sentence is 40 words long with three subclauses, none get pulled.
35

No thin pages outside utility routes

Pages under 300 words (tag archives, empty category pages, deprecated content) hurt site-level authority. Audit your low-content pages quarterly — noindex, consolidate, or remove anything that doesn't earn its slot.
36

Original quotes, data, or research appear on the page

AI engines reward originality. A page with a single original statistic, expert quote, or framework outranks ten pages summarizing the same secondary sources. Generate at least one citation magnet per pillar piece.
37

Evergreen content updated within the last 12 months

AI engines aggressively prefer recent content. Audit your top 20 evergreen pieces every quarter — update statistics, refresh examples, add new sections. Bump the dateModified. The freshness signal is real.

Pillar 06

Engine-Specific Coverage

Are you covering the quirks of each major AI engine?

38

ChatGPT — content is reachable on URLs Bing has indexed

ChatGPT search relies heavily on Bing's index for retrieval. If Bing hasn't indexed a page, ChatGPT can't surface it. Verify Bing indexation for every page you want cited.
39

ChatGPT — no authentication required

ChatGPT can't crawl content behind a login wall. Public versions of gated content (sales pages, marketing pages, public-facing docs) need to live at indexable URLs.
40

Perplexity — page loads in under 2 seconds without JS

Perplexity crawls aggressively but bails fast on slow pages. Target a sub-2-second First Contentful Paint without JavaScript execution. Static rendering is non-negotiable for Perplexity visibility.
41

Perplexity — content includes specific numbers, dates, and entities

Perplexity's ranking model rewards specificity. Pages with concrete numbers, dated events, and named entities get cited at meaningfully higher rates than abstracted explainers.
42

Google AI Overviews — content matches What/Why/How patterns

AI Overviews lift content that directly answers question-format queries. Structure pages around 'What is X?', 'Why X matters', 'How to X' sub-headings with self-contained 50–80 word answer blocks.
43

Claude — content sits on a site without aggressive bot blocking

Claude (and ClaudeBot) respect robots.txt rigorously. Most public sites are accessible by default, but if you've blocked at the CDN level (Cloudflare bot fight mode, etc.), Claude can't see your content.
44

Gemini — content has clear citations to authoritative sources

Gemini weights pages that themselves cite authoritative sources. Outbound links to government, academic, or major-publication primary sources signal that you're synthesizing rather than fabricating.

Pillar 07

Monitoring & Measurement

How do you know whether any of this is actually working?

45

Monthly prompt sweep against AI engines

Pick 20 prompts your buyers actually use. Run them against ChatGPT, Perplexity, Gemini, and Claude monthly. Record which brands appear, which URLs get cited, and how your visibility trends over time.
46

Competitive AI share-of-voice tracking

Track not just whether you appear, but how often you appear vs. competitors for the same prompts. AI share of voice is the metric your CMO actually cares about — define it and measure it monthly.
47

Referrer logs surface AI engine traffic

Filter your analytics for traffic from chat.openai.com, perplexity.ai, gemini.google.com, and claude.ai. The volume is small but rising quickly. Knowing which pages get cited is more actionable than knowing aggregate traffic.
48

Tracking tool subscription OR documented manual process

Profound, AthenaHQ, Otterly, and Peec AI productize this monitoring at $99–$500/month. If you don't subscribe, document a manual process — and assign owner, cadence, and reporting destination.
49

Monthly AI visibility report to leadership

If the C-suite doesn't see the data, AEO investment dies in the next budget cycle. Create a one-page monthly dashboard: share-of-voice trend, top-cited pages, top-competitor surfaces, gaps to close.
50

Quarterly comprehensive audit cadence

AEO best practices shift every quarter as models update. Schedule this 50-item audit as a recurring quarterly exercise. The teams treating AEO as a one-time project lose to the ones running it as an operating discipline.

Scoring

How to read your score.

Count one point for every item you can answer yes to. Items that are partially true count as zero — AEO rewards completion, not effort.

42 – 50: AEO-ready. You're in the top decile of websites. Most gaps are tactical, not structural. Focus on freshness and measurement.
32 – 41: Structurally sound. Discoverability and citability are likely fine. Authority signals and engine-specific coverage are the highest-leverage fixes.
21 – 31: Partially exposed. AI engines can find some of your content but not most. Schema, semantic HTML, and entity-rich writing are typically the gap.
0 – 20: Invisible. Your content is essentially unparseable to AI engines. Start with Pillar 01 (Discoverability) and Pillar 04 (Structured Data). Everything else is secondary until those land.

Keep reading

The AEO Audit Checklist

Public llms.txt file at /llms.txt

Robots.txt explicitly allows AI crawlers

Server response time under 800ms

Critical content renders without JavaScript

XML sitemap submitted to Google + Bing

Canonical URLs declared on every page

Pages return clean 200 status

Each page has exactly one H1

H2s and H3s use question-format phrasing

Lead paragraph answers the primary question in 50–100 words

Statistics and quotes are clearly attributed

Lists use semantic HTML, not styled divs

Tables use real <table> markup

Definitions use <dl><dt><dd> blocks

Author byline is visible with credentials

Author bio pages with credentials and external links

Comprehensive 'About' and 'Team' pages

Brand profiles on Wikipedia, Crunchbase, LinkedIn, industry directories

Backlinks from at least 3 high-authority domains

Publication date visible on every article

'Last updated' dates surfaced for evergreen content

Reviewed-by or fact-checked-by attribution where applicable

Organization schema on homepage and About page

Article schema on every editorial post

Author schema linked to Article schema

FAQPage schema on pages with question-format content

BreadcrumbList schema on category and sub-pages

Product or Service schema where applicable

All schema validates without errors

Each page targets a single primary intent

Substantive length — 800+ for posts, 1,500+ for pillars

Entity-rich writing — name people, companies, products explicitly

Internal links connect related entities and topics

Reading level appropriate to the audience

No thin pages outside utility routes

Original quotes, data, or research appear on the page

Evergreen content updated within the last 12 months

ChatGPT — content is reachable on URLs Bing has indexed

ChatGPT — no authentication required

Perplexity — page loads in under 2 seconds without JS

Perplexity — content includes specific numbers, dates, and entities

Google AI Overviews — content matches What/Why/How patterns

Claude — content sits on a site without aggressive bot blocking

Gemini — content has clear citations to authoritative sources

Monthly prompt sweep against AI engines

Competitive AI share-of-voice tracking

Referrer logs surface AI engine traffic

Tracking tool subscription OR documented manual process

Monthly AI visibility report to leadership

Quarterly comprehensive audit cadence

How to read your score.