This GEO audit checklist gives you 26 checks across five areas. Each check has a way to test it, a clear pass condition and a priority. Use it to find what stops ChatGPT, Gemini, Perplexity, Google AI Overviews and Google AI Mode from reading, trusting and citing your site, then fix the gaps that matter most. The stakes are real: AI Overviews has over 2.5 billion monthly users and AI Mode has passed one billion.
Key Takeaways
- Crawl access, indexing and schema are basics to confirm, and they don’t earn AI citations alone.
- Only 12% of URLs cited by AI assistants rank in Google’s top 10 for the same prompt.
- Google says Google AI Overviews and AI Mode need no special schema, markup or new files.
- The real audit is a baseline of mentions, citations, sentiment and sources on each AI engine.
- Fix in this order: access blockers, wrong facts, citation-source gaps, then page structure.
How to read this checklist: this GEO checklist works best on your 10-30 priority pages, not the whole site. High priority means the issue blocks citations or spreads a wrong fact. Medium means it lowers your odds of being cited. Areas 1-4 are basics to confirm alongside the audit. Area 5 is the audit itself.
GEO Audit Checklist: All 26 Checks at a Glance
The complete GEO audit checklist below lists every check with its area and priority. Each area section further down shows how to test the check and what a pass looks like.
| # | Check | Area | Priority |
|---|---|---|---|
| 1 | AI search crawlers are allowed in robots.txt | AI crawler access | High |
| 2 | Firewall and CDN let AI bots through | Crawler access | High |
| 3 | Priority pages are indexed | Crawler access | High |
| 4 | Main content is in the HTML | Crawler access | Medium |
| 5 | Nothing opts pages out of Google’s AI features | Crawler access | High |
| 6 | The answer comes first | Content | High |
| 7 | Headings match real questions | Content | Medium |
| 8 | Facts sit in extractable formats | Content | Medium |
| 9 | Claims are sourced and current | Content | High |
| 10 | One page answers one intent | Content | Medium |
| 11 | Organization schema identifies the brand | Schema and entity | Medium |
| 12 | Templates carry the right schema types | Schema and entity | Medium |
| 13 | Schema matches visible content | Schema and entity | Medium |
| 14 | Entity facts are consistent everywhere | Schema and entity | High |
| 15 | Authors are named and credible | Trust | Medium |
| 16 | Pages carry original evidence | Trust | Medium |
| 17 | The brand appears on domains AI cites | Trust | High |
| 18 | Third-party descriptions are accurate | Trust | High |
| 19 | Query Universe Coverage | Baseline | High |
| 20 | AI Visibility Score and Share of Voice | Baseline | High |
| 21 | Citation Source Forensics | Baseline | High |
| 22 | Sentiment and Brand Framing | Baseline | Medium |
| 23 | Content Type Mix | Baseline | Medium |
| 24 | Competitor Signal Analysis | Baseline | High |
| 25 | Commerce and PDP Readiness | Baseline | Medium (High for ecommerce) |
| 26 | Market and Language Coverage | Baseline | Medium (High for multi-market brands) |
What is a GEO Audit Checklist?
A GEO audit checklist is a fixed set of tests that shows whether AI answer engines can reach your pages, understand your brand, trust your claims and cite you. It also records how often they already do.
Generative engine optimization (GEO) is the practice of earning mentions and citations in AI answers. Answer engine optimization (AEO) is a closely related term, so a GEO checklist and an answer engine optimization checklist overlap heavily. GEO in this sense has nothing to do with geographic or local SEO.
A GEO audit also differs from a traditional SEO audit. An SEO audit asks where your pages rank. A GEO audit asks whether AI systems name and cite you, and which sources they trust instead. The full comparison is in our guide to the GEO audit vs SEO audit. This page is the checklist. For the step-by-step process, read how to run a GEO audit.
Can AI Crawlers Reach and Read Your Key Pages? (Checks 1-5)
AI crawlers can reach your key pages when search bots are allowed in robots.txt and your firewall, your priority pages are indexed, and nothing opts them out of AI features. These are technical basics to confirm alongside the audit, because AI systems can’t cite a page their crawlers never fetched.
Separate search bots from training bots before you change anything. OpenAI says sites that disallow OAI-SearchBot won’t be shown in ChatGPT search answers, and robots.txt changes take about 24 hours to apply. GPTBot collects content that may train OpenAI’s generative AI foundation models, so blocking it is a business decision, not a visibility fix. Anthropic splits ClaudeBot (training) from Claude-SearchBot (search indexing) the same way.
| # | Check | How to check | Pass looks like | Priority |
|---|---|---|---|---|
| 1 | AI search crawlers are allowed | Read the robots.txt rules for OAI-SearchBot, Claude-SearchBot, PerplexityBot, Googlebot and Bingbot | No user agent used for search is disallowed on priority pages | High |
| 2 | Firewall and CDN let AI bots through | Filter server logs by user agent and compare IPs with the ranges OpenAI and Perplexity publish | AI bots get 200 responses, not 403 errors or challenge pages | High |
| 3 | Priority pages are indexed | URL Inspection in Google Search Console and Bing Webmaster Tools, plus sitemap lastmod dates | Indexed, self-canonical and listed in the XML sitemap with accurate dates | High |
| 4 | Main content is in the HTML | “View crawled page” in URL Inspection, or fetch the raw HTML and search for key facts | Answers, prices and specs appear as text, not just in scripts, images or PDFs | Medium |
| 5 | Nothing opts pages out of Google’s AI features | Search Console Settings for the Search generative AI control, and page source for nosnippet or max-snippet | The setting is “Include” and no snippet limits sit on pages you want cited | High |
Check 2 is where hidden failures usually sit. Some CDN and bot-management settings block AI crawlers by default, so a clean robots.txt can still hide your site. OpenAI and Perplexity publish IP ranges for their crawlers, which lets your security team allowlist verified traffic instead of opening the door to every scraper. Perplexity recommends allowing PerplexityBot and its published IPs if you want to appear in its results.
Check 5 matters because Google added a Search Console control in 2026 that removes a site from AI-generated answers in Google’s AI Overviews, Google’s AI Mode and AI Overviews in Discover. Sites that opt out get no traffic or impressions from those features (Google). Confirm nobody switched it off.
Is Your Content Structured To Be Cited? (Checks 6-10)
Content is ready to be cited when each page answers its question in the opening lines, splits into sections that stand alone, and backs every claim with a dated source. AI tools prefer content they can quote without rewriting it, and AI answer engines reward pages that make that easy.
| # | Check | How to check | Pass looks like | Priority |
|---|---|---|---|---|
| 6 | The answer comes first | Read the first paragraph under the H1 on its own | It answers the main question in 40-60 words and names the subject | High |
| 7 | Headings match real questions | Compare H2s with the prompts buyers type into AI engines | Most H2s are questions, and each section makes sense read alone | Medium |
| 8 | Facts sit in extractable formats | Check where specs, prices, steps and comparisons live | They sit in tables, lists or text definitions with labelled rows and columns | Medium |
| 9 | Claims are sourced and current | Spot-check five statistics on each page | Each one links to a dated primary source, and the page shows an updated date | High |
| 10 | One page answers one intent | Run a site: search and review the Search Console query report | No two blog posts or comparison pages split the same query | Medium |
Good content structure helps human readers as much as AI engines. Direct answers at the top of a section let an engine lift one passage without misreading the page around it. Clear labels in tables stop it from confusing a price with a spec.
Check 10 is the one content teams skip. When multiple pages answer the same question, AI search systems have to pick one, and they may pick a weaker or older page. Merge overlapping pages before you rewrite them. Content quality improves faster from consolidation than from new posts.
Do Structured Data and Entity Facts Describe Your Brand Clearly? (Checks 11-14)
Structured data and entity facts describe your brand clearly when your schema markup matches the visible page and your brand reads the same on every site AI systems check. Google is direct on this point: there is no special schema.org structured data you need to add for Google AI Overviews or AI Mode.
Structured data is still useful. It helps machines classify a page, and Google lists matching it to visible text as a general practice that still applies. So audit structured data for accuracy, not volume. HowTo schema is a common distraction here. Google retired HowTo rich results in 2023, so a missing HowTo schema block is not a failure.
| # | Check | How to check | Pass looks like | Priority |
|---|---|---|---|---|
| 11 | Organization schema identifies the brand | Run the homepage through the Schema.org validator | Name, logo, URL and sameAs links match the About page and official profiles | Medium |
| 12 | Templates carry the right schema types | Test one URL per template in Google’s Rich Results Test | Article schema with a Person author on blog posts, Product schema on product pages | Medium |
| 13 | Schema matches visible content | Compare schema values with the text on the page | No errors, and no ratings, dates or claims that aren’t visible | Medium |
| 14 | Entity facts are consistent everywhere | Compare name, category and products across your own website, Google Business Profile, LinkedIn and directories | The same description everywhere, with no retired products or old positioning | High |
Check 14 carries the highest priority in this area because it shapes what AI engines say, not just what they read. If your About page calls you one thing and three directories call you another, engines may blend both. For implementation detail, see our guide to schema markup for AI search.
Do Trust and Off-Site Signals Back Up Your Brand? (Checks 15-18)
Trust and off-site signals back up your brand when named experts write your content, your pages carry evidence nobody else has, and the sites AI search engines cite describe you accurately. Ranking alone doesn’t decide citations. Ahrefs found that 80% of citations from AI assistants don’t rank in Google at all for the original query.
| # | Check | How to check | Pass looks like | Priority |
|---|---|---|---|---|
| 15 | Authors are named and credible | Check bylines, bio pages and Person schema | Every post has a named author with relevant experience and a linked bio | Medium |
| 16 | Pages carry original evidence | Count first-party data points, screenshots and case studies on priority pages | Each priority page has at least one fact nobody else can publish | Medium |
| 17 | The brand appears on domains AI cites | Take the top cited domains from check 21 and search each one for your brand | You are present on the review sites, publications and comparison pages that win citations | High |
| 18 | Third-party descriptions are accurate | Read your entries on review sites, directories and Wikipedia | Wrong facts are logged and corrected at the source | High |
Checks 17 and 18 often decide outcomes in competitive categories. If an AI answer cites a review roundup that leaves you out, no on-page fix changes that answer. Digital PR, analyst relations and partner content close these gaps, which traditional SEO link building often misses. Track every accurate brand mention on a cited domain. Treat AI citations of third-party pages as a target list, not a scoreboard.
Have You Measured an AI Visibility Baseline? (Checks 19-26)
An AI visibility baseline is the GEO audit itself: a fixed prompt set run on each AI engine to record where your brand appears, who gets cited and why. It is the measurable core of generative engine optimization (GEO). The eight checks below map to the eight dimensions of Addlly AI’s GEO Audit. Run each prompt more than once, on different days, because AI-generated responses vary between runs.
| # | Check | How to check | Pass looks like | Priority |
|---|---|---|---|---|
| 19 | Query Universe Coverage | Turn buyer questions from search, forums and reviews into a fixed prompt set | The prompt set covers the category’s real questions, not just branded ones | High |
| 20 | AI Visibility Score and Share of Voice | Run the prompts on all AI platforms and log mentions against named competitors | Mention rate and share of voice recorded per platform, never averaged | High |
| 21 | Citation Source Forensics | Log every cited URL in each AI answer and classify it | Each source is tagged owned, competitor-driven or third-party | High |
| 22 | Sentiment and Brand Framing | Read how each engine describes you | Tone and exact wording logged, with wrong or negative claims flagged | Medium |
| 23 | Content Type Mix | Group cited URLs by format | You know which formats (guides, reviews, product pages) earn citations | Medium |
| 24 | Competitor Signal Analysis | Run the same analysis for named rivals | You can explain where and why competitors are cited instead of you | High |
| 25 | Commerce and PDP Readiness | Check product prompts against your site and marketplaces | Product pages appear as sources with correct specs | Medium (High for ecommerce) |
| 26 | Market and Language Coverage | Run the prompt set in each market and language | Visibility recorded for every market you sell in | Medium (High for multi-market brands) |
AI visibility measurement depends on citation tracking, which turns this table into a trend. Record the date, engine, prompt and every cited URL, so next month’s run shows whether the brand appears more often and why. For metric definitions and reporting, see our guide to GEO and AEO KPIs.
Which first-party reports should sit alongside the baseline?
First-party reports confirm part of what the prompt baseline shows, but none covers every engine. The generative AI performance report in Google Search Console shows impressions from Google’s AI surfaces (Google AI Overviews and AI Mode) by page and country, and reached all sites worldwide on August 31, 2026. It shows no clicks.
Bing Webmaster Tools’ AI Performance report shows citations across Microsoft Copilot and AI summaries in Bing, with grounding queries. In analytics, track referral traffic from chatgpt.com and perplexity.ai next to organic traffic.
What Mistakes Make a GEO Audit Checklist Misleading?
A GEO checklist misleads when teams read single results as facts or treat technical passes as visibility. Four mistakes cause most of the damage:
- Grading a check off one run. Answer engines change their output between runs and days. Mark a baseline check as passed only when the result holds across repeat runs.
- Blocking search and training bots with one rule. A blanket rule to block AI crawlers removes you from AI search results along with training datasets. Decide on each user agent separately.
- Averaging every engine into one score. AI platforms cite differently. Ahrefs found 28.6% of Perplexity’s cited URLs rank in Google’s top 10, against about 8% for the other assistants it tested.
- Treating schema as a ranking factor. Passing checks 11-13 makes your facts easier to read. It doesn’t put you in AI answers on its own.
How Do You Prioritize Your Fixes?
Fix what blocks access or spreads wrong facts first, then close citation gaps, then improve page structure. This order puts the cheapest, highest-impact fixes ahead of slow content work:
- AI crawler access blockers (checks 1-3 and 5). One robots.txt line or firewall rule can hide important pages from every engine. These fixes are usually quick.
- Wrong facts (checks 14, 18 and 22). Correct the source, not just your own site, because engines repeat what third parties say.
- Citation-source gaps (checks 17, 21 and 24). Win a place on the domains that already get cited for your category.
- Content structure and evidence (checks 6-10, 15 and 16). Rewrite priority pages so engines can lift answers and trust them.
- Rendering and schema hygiene (checks 4 and 11-13). Clean these up during normal content optimization and template work.
Re-run the baseline after each batch so you can tie changes in AI visibility to specific fixes. That link is how generative engine optimization proves its value to finance.
How Does Addlly AI’s GEO Audit Cover This Checklist?
Addlly AI’s GEO Audit runs Area 5 in full and uses the results to rank fixes for the rest of the checklist. Addlly AI is an AI search visibility platform for enterprise and multi-market brands. Its GEO Audit runs 250+ consumer-style prompts per brand across ChatGPT, Gemini, Perplexity, Google AI Overviews, Google AI Mode, Claude and Microsoft Copilot. It reports results per platform, traces the source behind every answer, and benchmarks named competitors. The output is a prioritized action plan ranked by commercial impact.
The GEO Audit is a knowledge and citation audit, not a technical SEO crawl, so confirm areas 1-3 with your SEO team. Fixes then run through Addlly’s GEO agents or its Managed GEO Services team. Addlly AI is SOC 2 Type II compliant and ISO 27001 certified.
Frequently Asked Questions About GEO Audit Checklist
How Often Should You Run a GEO Audit?
Run the AI visibility baseline every month with the same fixed prompts, so results stay comparable from one run to the next. Recheck crawler access and areas 2-4 after any site migration, redesign, CDN change or robots.txt edit, since one of those can undo months of work in a single release.
Can ChatGPT Do a GEO or SEO Audit for You?
Partly. ChatGPT can review a pasted page for structure, clarity and missing direct answers. It can’t see your server logs or Search Console data, or how other AI platforms answer your buyers across repeated runs. So it can review a page, but it can’t produce a baseline or a share-of-voice figure.
How Many Prompts Should a GEO Audit Use?
A GEO audit should use enough prompts to cover the real questions in your category, across every market you sell in. A first manual pass of a GEO checklist can start with 20-50 high-intent prompts. Addlly AI’s GEO Audit uses 250+ consumer-style prompts per brand, drawn from forums, reviews and social discussion.
Do You Need llms.txt To Pass a GEO Audit?
No. llms.txt is an optional, proposed file that points AI tools to key pages. Google says Google’s AI features need no new machine-readable files. Treat it as low-cost hygiene, not a pass condition. Learn more about what llms.txt is.
What Does a GEO Audit Produce?
A GEO audit produces a per-engine picture of how each AI answer engine treats your brand: mention rate, share of voice, sentiment, the sources behind each answer and how you compare with named competitors. A useful audit ends with a prioritized fix list, so each finding has an owner and a next step.
Get The GEO Audit of Your Website
This checklist shows you what to test. Generative engine optimization only pays off when the fixes ship. Addlly AI shows you what AI engines actually say about your brand and fixes the gaps that cost you citations.
- GEO Audit: a per-platform visibility baseline with citation source forensics and a ranked action plan.
- AI GEO Agent: creates and updates AI-ready content in your brand’s voice.
- AI Schema Markup Generator: generates structured data for your key page templates.
- PDP AI Agent: tunes product pages for AI shopping answers and marketplaces.
