A GEO audit checklist is a structured, step-by-step framework for testing whether your website is discoverable, extractable, and citable by AI search engines, including Google AI Overviews, ChatGPT, Perplexity, and Claude. The framework evaluates four layers: technical crawl access for AI bots, content citation-readiness, entity clarity, and E-E-A-T trust signals. Unlike a traditional SEO audit focused on keyword rankings, an AI search readiness audit measures whether generative engines can retrieve, understand, and confidently quote your content in AI-generated answers.

What follows is the full 40-point GEO audit checklist Passionfruit uses for B2B SaaS and consumer brands. The checklist is sorted by the four audit layers, with the test pattern and the fix for each item. Run it page by page or as a site-wide audit. Total time for a small site is around six hours; for an enterprise site, plan two to three weeks.

What is a GEO audit and why it differs from a traditional SEO audit

A GEO audit is the work of testing whether AI search engines can find, parse, and cite your content. The audit measures presence in AI-generated answers, not ranking position in blue-link results. The audit weighs signals AI systems use during retrieval: entity clarity, source crediting, content structure, freshness, and crawl access for the bots AI platforms use.

The shift matters because AI search now drives a real share of discovery. Conductor's January 2026 analysis of 21.9 million queries found AI Overviews appear on 25.11% of Google searches, while Alphabet's Q4 2025 earnings disclosure put AI Mode at 75 million daily active users. SEMrush's September 2025 research showed roughly 92 to 94% of AI Mode searches end without an external click, which makes citation share, not ranking position, the practical variable for visibility on these surfaces. Traditional SEO audits don't measure citation share, AI bot crawl access, or entity sorting, so they miss most of what determines GEO performance.

For the broader framework that this checklist plugs into, Passionfruit's generative engine optimization guide covers the strategic context across ChatGPT, Perplexity, Gemini, Claude, and Copilot.

Layer 1: Technical crawl access for AI bots

Technical crawl access is the first layer. Content that can't be reached by AI crawlers can't be cited. Most teams set up robots.txt and indexation for Googlebot and stop there, missing the dedicated bots AI platforms now run, which often have different parsing behavior and JavaScript limitations.

1. AI bot access in robots.txt

Test: Open /robots.txt and search for each of these user-agents: GPTBot, ChatGPT-User, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, Applebot-Extended, Amazonbot, Bytespider, CCBot. Confirm none are blocked unless you intend to block them.

Fix: Remove Disallow: / rules for AI bots you want indexing your content. If you want to allow AI search citation but block AI training, allow OAI-SearchBot, ChatGPT-User, PerplexityBot, and ClaudeBot while blocking GPTBot, Google-Extended, Applebot-Extended, and CCBot.

2. CDN and firewall rules

Test: Run a curl request with each AI bot user-agent against a sample page and confirm a 200 response, not a 403 or challenge page. Cloudflare's default "AI scrapers and crawlers" block setting blocks several AI bots without site-owner awareness.

Fix: In Cloudflare, navigate to Security → Bots and confirm AI bot blocking is configured purposefully. In other CDNs, check WAF rules for user-agent based blocking.

3. Server-side rendering of critical content

Test: Disable JavaScript in your browser and reload key pages. Confirm headings, body copy, structured data, and meta tags are all visible without JavaScript execution.

Fix: Most AI crawlers including OAI-SearchBot and PerplexityBot do not well execute JavaScript. Move content from client-side hydration to server-side rendering. For Next.js, use getServerSideProps or static generation. For Astro and similar frameworks, server rendering is the default.

4. JSON-LD delivery method

Test: View page source (not the rendered DOM) and confirm <script type="application/ld+json"> blocks appear in the raw HTML.

Fix: Server-render JSON-LD blocks. Don't inject them client-side through Google Tag Manager or React effects. Schema injected post-hydration is invisible to most AI crawlers.

5. Sitemap freshness

Test: Open /sitemap.xml and check <lastmod> timestamps. Confirm all are within the past 90 days for content that has been updated.

Fix: Configure your CMS to update sitemap <lastmod> on its own when a page is edited. Use IndexNow for instant push to Bing and Copilot, which Microsoft's Fabrice Canel confirmed at SMX Munich 2025 is a meaningful freshness signal for Copilot.

6. Page speed for AI parsing

Test: Run Google PageSpeed Insights and confirm Largest Contentful Paint is under 2.5 seconds and Interaction to Next Paint is under 200ms.

Fix: Compress images, defer non-critical JavaScript, and serve critical CSS inline. Pages that fail Core Web Vitals are less likely to be fully parsed by AI crawlers within their time budget.

7. Mobile-first rendering

Test: Use Google's Mobile-Friendly Test and confirm pass status. Verify content parity between mobile and desktop versions.

Fix: AI crawlers often use mobile rendering. Eliminate any content gated to desktop-only or hidden on mobile breakpoints.

Layer 2: Content citation-readiness

Content citation-readiness is the second layer. Crawl access without extractable content earns nothing. AI engines prefer content with definition-first paragraphs, predictable structure, current data, and explicit sourcing.

8. Definition-first paragraph openings

Test: Read the first sentence of each major section heading. Confirm it answers the heading's implicit question directly, without preamble.

Fix: Rewrite section openings to lead with a standalone definition. Growth Memo's February 2026 research on citation spread found 44.2% of LLM citations come from the first 30% of an article's text. Front-loading the strongest claims matters.

9. Section length and rhythm

Test: Count words between subheadings on key pages. Confirm sections fall between 120 and 180 words.

Fix: Break sections under 50 words by adding context; break sections over 250 words by splitting them. SE Ranking's November 2025 work found pages with 120-180 words between subheadings earn 70% more citations than pages with sections under 50 words.

10. Sentence length cap

Test: Run your content through a readability tool such as Hemingway or textstat. Confirm average sentence length is under 18 words.

Fix: Split long sentences. AirOps research from April 2026 found pages averaging 10 or fewer words per sentence earn 18.8% more AI citations than longer-sentence pages. Aim for 12 to 15 words on average.

11. Comparison tables on commercial pages

Test: On "best X" or "compare Y" pages, confirm at least one comparison table is present.

Fix: AirOps research found comparison pages with three tables earn 25.7% more citations than equal pages without comparison structure. Add tables for feature comparisons, pricing tiers, and use case matching.

12. List sections on validation pages

Test: On listicles, validation pages, and benefit-focused content, count discrete list sections.

Fix: AirOps research found validation pages built around eight list sections earn up to 26.9% more citations. Re-shape dense prose into named, scannable list sections.

13. Explicit sourcing with primary citations

Test: For every statistic, study reference, or named claim, confirm the source is hyperlinked to the original publisher (not an aggregator blog).

Fix: Swap secondhand citations for primary source links. AI engines weight pages with explicit, traceable sourcing higher when picking citations. Industry blogs that cite other industry blogs without primary anchors get downweighted.

14. Data point density

Test: Count specific data points (numbers, dates, named studies, attributed quotes) per 500 words. Aim for at least 3 per 500 words on informational content.

Fix: Add statistics, dates, and named source creditings where claims now sit unsupported. Generic claims like "many users report" should become "Pew Research's July 2025 study found 8% click rates when AI summaries appeared".

15. FAQ section presence on relevant pages

Test: On topic-pillar and how-to pages, confirm an FAQ section exists with 4-8 question-and-answer pairs.

Fix: Add an FAQ section answering the "people also ask" patterns for your target query. The Schanbacher 2026 study, published in the Journal of Advance Research in Business, Management and Accounting, found FAQPage schema correlated with much higher ChatGPT visibility across 1,508 sites.

16. Visible last-updated date

Test: Confirm a visible "last updated" or "reviewed" date appears at the top or end of every editorial page.

Fix: Show dateModified in the visible content, not just in schema. Freshness signals operate on both layers, and AI engines reading visible dates assess content currency independently of metadata.

17. Question-shaped subheadings

Test: Count subheadings phrased as questions ("What is...", "How does...", "Why does..."). Confirm at least 30% of subheadings are question-shaped on informational content.

Fix: Convert flat subheadings to question form where natural. Question-shaped headings match the everyday language AI engines hear from users.

18. Author crediting and bio

Test: Check that every article has a named author, an Author schema block, and a linked bio page with credentials.

Fix: Add author profiles with credentials, areas of expertise, and external links to verified profiles (LinkedIn, professional registries). Author signals feed E-E-A-T weighting in AI engine source picks.

Layer 3: Entity clarity

Entity clarity is the third layer. AI engines need to sort out who you are and what you do before they can cite you. Pages with sharp entity signals get picked over equivalent pages with fuzzy ones.

19. Organization schema with sameAs links

Test: Inspect Organization schema on the homepage and confirm a sameAs array linking to verified external identifiers: Wikidata, Crunchbase, LinkedIn company page, official social profiles.

Fix: Add the full sameAs array. The Wikidata link is the most important single field because AI engines use Wikidata as an entity resolution anchor.

20. Wikidata entry

Test: Search Wikidata for your brand. Confirm an entry exists with consistent naming, founding date, headquarters location, and official URL.

Fix: Create a Wikidata entry if missing. Wikidata is open-edit but moderated; create the entry, add citations to your own About page and to coverage in set up launchs, and link to your other verified identifiers.

21. Knowledge Graph presence

Test: Google your brand name. Confirm a Knowledge Panel appears on the right side of the SERP with logo, founding year, and headquarters location.

Fix: If absent, submit Knowledge Panel checking through Google. Ensure your About page is rich in entity-defining content (founding story, named founders, headquarters, registration details).

22. Consistent NAP across the web

Test: Search for your brand's name, address, and phone number across Google Business Profile, Apple Maps, Bing Places, Crunchbase, LinkedIn, and major industry directories. Confirm all five fields match exactly.

Fix: Update mismatches at the source. NAP consistency is checked by AI engines when validating local and regional citations.

23. Person schema for key contributors

Test: On articles by named authors, executive bios, and case study quotes, confirm Person schema is in place with jobTitle, worksFor, and sameAs to LinkedIn.

Fix: Add Person schema to author bios and quoted experts. AI engines that cite individual claims often need to verify the person before surfacing the citation.

24. Internal entity linking

Test: For every distinct concept introduced on a page (product names, methods, frameworks), confirm at least one internal link to the page that defines it.

Fix: Build a topic graph and link every concept mention to its canonical definition page. AI engines reading internal link patterns judge topical authority from this structure.

25. Topic clusters with pillar pages

Test: Identify your top three topics. Confirm each has a pillar page linked from at least 5-8 supporting articles using descriptive, varied anchor text.

Fix: If the cluster doesn't exist, build it. If anchor text repeats ("click here", "read more"), rewrite to clear 3 to 5 word phrases that name the concept.

26. Brand mention consistency

Test: Audit how your brand name shows up across owned content. Confirm consistent capitalization, spacing, and formal name versus shorthand.

Fix: Pick one canonical brand spelling and update all instances. Variations confuse AI entity resolution.

Layer 4: E-E-A-T trust signals

E-E-A-T trust signals are the fourth layer because even discoverable, extractable, entity-clear content can be skipped if AI engines don't trust the source. Trust signals are how AI engines pick which of several valid sources to cite.

27. About page depth

Test: Read your About page. Confirm it covers the founding story, leadership, headquarters location, registration details, and how the company is funded.

Fix: Expand the About page to cover all five. AI engines reading About pages assess source credibility from how much checkable detail is present.

28. Author credentials prominently displayed

Test: On every article, confirm the author's name, role, and credentials appear above the fold and link to a full bio page.

Fix: Add credential lines under the author byline. "12 years in B2B SaaS marketing, earlier at Acme" reads as a trust signal. "Content Writer" alone does not.

29. Editorial policy page

Test: Confirm a publicly accessible editorial policy or content guidelines page exists, covering review process, sourcing rules, and correction policy.

Fix: Publish an editorial policy at /editorial-policy or /content-standards. AI engines weight content from sites with explicit standards higher than content from sites without.

30. Reviewed-by signals for sensitive topics

Test: On health, finance, legal, or other high-stakes topics, confirm a "Reviewed by [expert name]" line appears with credentials and a reviewer bio link.

Fix: Add expert review for sensitive content. For YMYL (Your Money or Your Life) topics, this signal moves citation eligibility from contested to default.

31. Citation outbound to peer-reviewed and primary sources

Test: Count outbound links to .gov, .edu, peer-reviewed journals, or named set up launchs per 1000 words. Aim for at least 2 per 1000 words on informational content.

Fix: Replace blog-aggregator citations with primary source links. AI engines that follow citation chains backward stop at primary sources. Pages citing only secondary sources get less weight.

32. Original research or first-party data

Test: On at least 10-20% of your blog content, confirm the inclusion of original survey data, internal benchmark studies, or first-party customer insights.

Fix: Run quarterly micro-surveys or analyses against your own customer data and publish the findings. Pages with original data tend to become cited sources for other publishers. That builds entity authority.

33. External citations of your brand

Test: Search for "site:non-yourdomain.com 'your brand name'" and count external mentions across the past 6 months. Confirm coverage in at least 3 set up industry launchs.

Fix: Run PR and outreach where the coverage gap is largest. External mentions in trusted launchs validate your brand entity for AI engines.

34. Trust signals in the footer

Test: Confirm the footer contains a physical address, contact email, privacy policy link, and terms of service link.

Fix: Add missing trust signals to the global footer. AI engines that crawl site-wide elements look for these signals as baseline credibility markers.

35. Security and HTTPS

Test: Open the site in a browser and confirm an HTTPS padlock appears. Run an SSL Labs scan and confirm a grade of A or higher.

Fix: Renew or upgrade certificates. Insecure sites get downweighted across both traditional and AI search.

Layer 5: Measurement (the audit cycle)

Measurement is the layer most audits skip. Without a tracking baseline, you can't tell whether the fixes worked. The checks below establish the measurement layer.

36. AI citation tracking baseline

Test: Pick 10 brand-relevant prompts. Run each three times against ChatGPT, Perplexity, Gemini, and Claude. Log which brands appear and in what order.

Fix: Save the baseline. Re-run quarterly. SparkToro's January 2026 research found fewer than 1 in 100 paired runs of the same prompt return the same brand list, so repeated sampling is the minimum reliable method. Passionfruit's research on AI brand recommendation variability covers the math behind why single-run tracking misses most of the signal.

37. Search Console AI feature performance

Test: Open the Search Console Performance report and filter to Web search type. AI Overview and AI Mode traffic is included here, not broken out separately.

Fix: Build a comparison view of impressions and clicks across quarters. Important caveat: Google confirmed on April 3, 2026 that Search Console impression data was inflated by a logging bug from May 13, 2025 through April 27, 2026. Year-over-year compares spanning that window need to be flagged. Passionfruit's research on Search Console measurement trust covers what this means in practice.

38. Branded versus non-branded segmentation

Test: In Search Console, apply the branded queries filter (rolled out fully on March 11, 2026) and confirm branded versus non-branded splits are visible.

Fix: Track the two segments separately. Branded growth signals brand recall lift from AI citation. Non-branded growth signals retrieval gains.

39. Server log analysis for AI bot crawl frequency

Test: Pull server logs from the past 30 days and count requests by user-agent for each AI bot identified in step 1.

Fix: If AI bot requests are low or absent, revisit step 1 and step 2. Search Engine Journal reported in April 2026 that ChatGPT-User now generates 3.6x more crawl requests than Googlebot in many datasets, so absence of AI bot crawl activity is a strong signal something is blocking them.

40. Brand tracking platform rollout

Test: Confirm a tool such as Profound, Otterly, or Passionfruit Labs is configured to track brand citations across AI platforms.

Fix: If absent, deploy one. Manual tracking does not scale past 20-30 prompts. Passionfruit's guide to brand tracking across ChatGPT, Perplexity, and Google AI covers the practical setup.

How long does a GEO audit take?

A small site (under 50 pages) with set up SEO basics: roughly 6 hours of working time spread across 1-2 days. Layer 1 (crawl access) and Layer 5 (measurement baseline) take half the time. A mid-size site (50-500 pages): 1-2 weeks, with some checks (content rewriting, schema deployment) extending into the following month. An enterprise site (500+ pages): 2-3 weeks for the audit phase, then a phased fix-up plan running over the following quarter.

The audit is worth running on at least an annual cadence. AI engines change retrieval behavior often enough that the schema and structural targets shift on roughly a 6-12 month cycle. Ahrefs published a March 2026 update showing AI Overview citation overlap with Google's top 10 dropped from 76.1% to 38% in eight months. The drop was driven partly by AI Overviews moving to Gemini 3 in January 2026. Static checklists go stale.

For the schema rollout side of the audit, Passionfruit's guides to AI-friendly schema markup and FAQ schema for AI answers cover the code patterns for items 18, 23, and 24.

Common GEO audit mistakes

Five mistakes drive most failed GEO audits. The list: auditing only the technical layer, treating the audit as one-time work, conflating ranking with citation, skipping the measurement baseline, and over-optimizing without first checking access.

Auditing only the technical layer (Layer 1) is the most common failure. Technical fixes are necessary but not enough on their own. A site with perfect crawl access and weak content gets no citations. A site with strong content and blocked crawlers gets no citations either. Both layers compound.

Treating the audit as one-time work ignores how fast the underlying AI engines move. The 8-month drop in AIO/top-10 overlap is one example. The Gemini 3 switch is another. Quarterly mini-audits on the same checklist catch drift before it builds into a ranking loss.

Conflating ranking with citation produces wrong fixes. Better blue-link rankings do not on its own translate to more AI citations. SE Ranking's August 2025 work found only 14% of URLs cited in AI Mode also rank in Google's top 10 for the same query. The two surfaces require different optimization moves, even when the underlying content is shared.

Skipping the measurement baseline means the audit fixes can't be evaluated. Without a quarter-one baseline, the quarter-two numbers are just numbers. Run step 36 before any other fix-up work.

Over-optimizing without first checking access (Layer 1) is the silent killer. Teams that spend three weeks rewriting content for AI pulling and only then discover GPTBot is blocked in robots.txt have wasted the work. Layer 1 first, every time.

Run your GEO audit with Passionfruit

Running a full GEO audit at scale takes a system, not just a checklist. Brands moving from ad-hoc AI search work to consistent citation visibility usually need help with the audit phase, the schema and content fix-up phase, and the AI citation measurement layer that sits outside Search Console. To run a GEO audit that produces consistent AI citation across ChatGPT, Perplexity, Gemini, and Google AI Overviews, start with Passionfruit Labs for self-serve AI visibility tracking, explore the end-to-end AI search and SEO growth service, or request a quote. The Passionfruit case studies show the full GEO audit framework applied across B2B SaaS and consumer brands.

Frequently asked questions

What is a GEO audit checklist?

A GEO audit checklist is a structured, step-by-step framework for testing whether your website is discoverable, extractable, and citable by AI search engines, including Google AI Overviews, ChatGPT, Perplexity, and Claude. The checklist evaluates four layers: technical crawl access for AI bots, content citation-readiness, entity clarity, and E-E-A-T trust signals.

How is a GEO audit different from an SEO audit?

Old-school SEO audits measure keyword rankings, backlink profiles, and on-page elements that drive blue-link visibility. GEO audits measure citation share in AI-generated answers, AI bot crawl access, entity sorting, and structural extractability. The two overlap on technical SEO basics but diverge on what counts as a win. Citation share is the GEO outcome; ranking position is the traditional SEO outcome.

How often should I run a GEO audit?

Annually for the full 40-point audit, quarterly for a focused mini-audit on Layer 1 (crawl access) and Layer 5 (measurement). AI engines change retrieval behavior often enough that the schema and structural targets shift on roughly a 6-12 month cycle, so static yearly audits miss meaningful drift.

Which AI platforms should I prioritize in a GEO audit?

Audit for the platforms your audience actually uses. Statcounter's April 2026 data placed ChatGPT at 78.16% of total AI chatbot referrals, Google Gemini at 8.65% (overtaking Perplexity), and Anthropic's Claude at 2.91%. For most B2B and consumer audiences, ChatGPT and Google AI Overviews carry the most weight, with Gemini and Perplexity second-tier and Claude growing fast.

What schema types matter most for GEO audits?

The schema types with the highest citation-pulling impact are FAQPage, Article and BlogPosting, Organization (with full sameAs links), LocalBusiness for region-specific pages, and HowTo for step-by-step content. The Schanbacher 2026 study found FAQPage schema was the strongest single predictor of ChatGPT visibility in a sample of 1,508 sites.

Can I run a GEO audit myself or do I need an agency?

A small site under 50 pages can usually be audited in-house in 6 hours of working time, given basic SEO familiarity. Mid-size and enterprise sites benefit from agency support for the schema deployment, content fix-up, and measurement setup steps. The audit itself is checkable against this article; the fix-up work scales with site complexity.

What's the most important step in a GEO audit?

Layer 1, step 1 (AI bot access in robots.txt) is the most important step because everything downstream depends on it. A site with a perfect content layer and blocked AI bots earns zero AI citations. Check robots.txt and CDN rules first, every time.