Generative engine optimization (GEO) audit checklist
Run our free SEO audit tool to check your site against these 200 criteria automatically.
Written by Vasa, GEO & AEO specialist · Last updated
Work through it rather than read it: every check is tickable and your progress is saved in this browser, so you can stop and come back. Each one carries how badly it hurts to skip it (13 are critical), roughly how long the fix takes, and whether it is a developer job or a content job, so you can filter down to what you can ship today and hand the rest over as tickets.
Start here: the checks that keep you out of AI answers entirely
0 of 9 cleared
Most GEO work shifts how often you get cited. These change whether you can be cited at all — an answer engine that cannot fetch or attribute the page will never quote it, however well written it is.
- — one wildcard disallow removes you from every answer engine at once details
- — block either agent and ChatGPT can never quote the page, however good it is details
- — Perplexity crawls on its own, so a block there costs you its citations alone details
- — Claude fetches under user agents most robots.txt files never mention details
- — disallowing it opts you out of AI Overviews and Gemini while you still rank details
- — content that only appears after JavaScript runs is invisible to most AI fetchers details
- — an answer spread over six paragraphs never gets lifted as a passage details
- — with no entity in structured data, the answer gets attributed to someone else details
- — an unsourced number is the first claim an answer engine drops details
Generative engine optimization audit vs traditional SEO
Classic SEO competes for a ranked link, so it optimises the page as a whole. Generative engines retrieve and quote fragments, so they reward pages made of self-contained, attributable passages. A page can rank on page one and still never be cited, usually because its answer is spread across six paragraphs, its crawler access is blocked for AI user agents, or its content only exists after JavaScript runs.
GEO readiness audit: what makes a site AI-search ready
A complete GEO site audit evaluates technical accessibility alongside semantic authority. If an AI agent cannot fetch your raw markdown or HTML without client-side hydration, or cannot attribute a statement to an authoritative entity, your page is skipped in favor of sites that provide structured, verifiable answers.
AI search compatibility check
AI search compatibility is the pass-or-fail layer of generative search optimization: can GPTBot, PerplexityBot and ClaudeBot fetch the page, is the answer in the server HTML, and does the page say who is behind it. If any of these fails, no amount of content work gets you cited. The free audit tests crawler access per user agent.
GEO content audit: re-checking pages you already have
A GEO content audit runs the answer, chunk and evidence checks over pages that already rank, to find the ones AI engines skip. Start with pages that get search traffic but no AI citations: they usually bury the answer below the first screen, split it across paragraphs that lean on each other, or quote statistics without a source. Rewriting the opening passage is often the whole fix.
How to do a GEO audit
The step-by-step workflow, in the order that matters.
- 01Inspect AI crawler access in robots.txt
Verify that robots.txt allows AI crawlers including GPTBot, PerplexityBot, ClaudeBot, and Google-Extended to fetch your pages.
- 02Format answers for first-passage quotability
Deliver your direct answer in the opening 40 to 60 words using clean, server-rendered HTML rather than client-side JavaScript.
- 03Structure content into semantic chunks
Break complex sections into question-shaped headings, comparison tables, and self-contained paragraphs that models can cite independently.
- 04Ground your brand and authors with entity schema
Implement Organization and Person JSON-LD schema with verified sameAs profiles on Wikipedia, LinkedIn, and official directories.
- 05Add verifiable citations and publish llms.txt
Include primary data citations and create an /llms.txt file to guide AI crawlers and developer agents to your highest-value content.
AI crawler access
0 / 8 done
An answer engine can only cite what it can fetch. This is the single most common reason a good page never appears in AI answers.
- Critical≤ 5 minDev
Whether robots.txt allows OpenAI's crawler and search fetcher on this path.
- Critical≤ 5 minDev
Perplexity crawls and cites separately from Google.
- Critical≤ 5 minDev
Anthropic's crawlers evaluated against your rules.
- Critical≤ 5 minDev
Controls Gemini and AI Overviews grounding independently of Googlebot.
- Important≤ 5 minDev
Two more retrieval paths most sites never check.
- Critical≤ 5 minDev
A wildcard disallow that silently removes you from every answer engine.
- CriticalHalf a day+Dev
Client-rendered content is frequently invisible to AI fetchers.
- Important≤ 5 minDev
Served as parseable HTML with a declared charset.
Answer readiness
0 / 8 done
AI engines lift self-contained passages. This group scores whether such passages exist on your page at all.
- Important~30 minContent
The tool extracts the passages most likely to be quoted and shows them to you.
- Critical~30 minContent
The core question is answered within the first 40 to 60 words.
- Important~30 minContent
Headings phrased the way people actually ask.
- Important~30 minContent
Passages that still make sense with all surrounding context removed.
- Important≤ 5 minContent
'X is …' constructions models reuse verbatim.
- Important~30 minContent
Structured formats are extracted far more reliably than prose.
- Important~30 minContent
Depth of the first substantive answer in the document.
- Polish~30 minContent
Passages leaning on 'it' and 'this' cannot be quoted standalone.
Chunkability and structure
0 / 6 done
Retrieval works on chunks. Badly structured pages chunk badly and lose to worse content.
- Important~30 minContent
Sections short enough to survive chunking, long enough to be complete.
- Important~30 minContent
Enough headings that each chunk carries its own topical label.
- Polish~30 minDev
article, section and main instead of nested divs.
- Important≤ 5 minDev or content
Skipped levels break the chunk-to-topic mapping.
- PolishHalf a day+Dev
How much of the page is nav and footer rather than content.
- Polish~30 minDev or content
An in-page TOC helps both users and retrieval systems.
Entity clarity
0 / 6 done
Models must know who you are before they will attribute an answer to you.
- Important≤ 5 minContent
One canonical name, not five variations across the page.
- Critical~30 minDev
The entity exists in structured data, not just prose.
- Important≤ 5 minDev
Profiles that let a model disambiguate you from similarly named entities.
- Important≤ 5 minContent
Attribution depends on the brand appearing near the answer.
- Important~30 minContent
What you offer, stated in plain language a model can repeat.
- Important≤ 5 minDev or content
Drift between markup and text weakens the entity.
Evidence and trust
0 / 8 done
Answer engines prefer sources that look verifiable. These are the signals they can see.
- Important~30 minContent
Concrete numbers are the most-quoted content type in AI answers.
- Critical~30 minContent
An unsourced number is a liability, not an asset.
- Important~30 minContent
Links to the origin of each claim.
- Important~30 minContent
Authoritative sources versus self-referential loops.
- Important≤ 5 minDev or content
Freshness is a hard filter for many AI retrieval systems.
- Polish≤ 5 minDev
Visible dates matching schema dates.
- ImportantHalf a day+Content
Content that cannot be paraphrased from elsewhere.
AI-native files and interfaces
0 / 6 done
The emerging layer. Very few sites have any of it, which makes it cheap differentiation.
- Important≤ 5 minDev or content
Structure validated against the llmstxt.org proposal.
Generate an llms.txt - Polish~30 minContent
Real page inventory rather than a marketing paragraph.
- Polish~30 minDev
Full-content variant for documentation-heavy sites.
- PolishHalf a day+Dev
Whether assistants can call your site as a tool, not just read it.
- Polish~30 minDev
RSS, JSON feeds or an API a model can follow.
- Important~30 minDev
FAQPage, HowTo, Article and Organization where they genuinely apply.
AI-generated content
0 / 8 done
Drafting with AI is fine; publishing it unchecked is not. Google holds AI-assisted pages to the same accuracy and usefulness bar as any other page, and treats mass-produced pages that add nothing as spam. No crawler can tell whether a person checked the facts, so these checks are yours to make.
Source: Google Search's guidance on generative AI content
- Critical~30 minContent
Someone who knows the subject reads and edits the draft before it goes live. A spell-check pass is not a review.
- Critical~30 minContent
Models predict likely words rather than look facts up, so every number, date and claim is checked against its source before publishing.
- Critical≤ 5 minContent
Each citation exists, opens, and supports the sentence it is attached to. A hallucinated reference costs more trust than having none.
- Important≤ 5 minDev or content
AI-written title tags, meta descriptions and image alt text appear in search results, so they get the same review as the body copy.
- Important≤ 5 minDev
Generated JSON-LD describes only what is visible on the page and follows Google's structured data policies.
- CriticalHalf a day+Dev or content
Many near-identical pages generated to catch search variations, templated doorway pages included, can breach Google's scaled content abuse policy.
- Polish≤ 5 minContent
A short note on how automation was used, when that context matters to the reader. Google suggests it; it is not required.
- Important~30 minDev
For ecommerce: Merchant Center requires AI-generated images to carry IPTC DigitalSourceType TrainedAlgorithmicMedia metadata, and AI-generated product titles and descriptions to be labeled as such.
Structured data completeness audit for GEO and AEO
A structured data completeness audit checks that every Schema.org block on a page parses and carries the properties its type needs. A missing required property is an error that makes the block unusable. A missing recommended property is a warning: the block still parses, but answer engines are left guessing about the author, the brand or the price.
These are the rules the IMVASA audit applies to JSON-LD and microdata on every page it scans.
| Schema type | Required (error if missing) | Recommended (warning if missing) |
|---|---|---|
| Organization | name | url, logo, sameAs |
| Person | name | url, jobTitle, sameAs |
| Article | headline | author, datePublished, image |
| FAQPage | mainEntity | — |
| HowTo | name, step | totalTime, image |
| BreadcrumbList | itemListElement | — |
| Product | name | image, description, offers, brand |
| Offer | price, priceCurrency | availability, url |
| AggregateRating | ratingValue, reviewCount | bestRating |
| Review | itemReviewed, reviewRating, author | — |
| LocalBusiness | name, address | telephone, openingHours, geo, sameAs |
| VideoObject | name, thumbnailUrl, uploadDate | description, duration |
Completeness is the floor, not the ceiling. For GEO, the properties that matter most are the ones that tie a claim to an entity: author and dates on articles, sameAs on Organization and Person, brand and offers on products. The audit also follows @id references across blocks and flags any that point to a node the page never defines. To write valid blocks from scratch, use the schema markup generator.
How to run a GEO audit for clients
To run a GEO audit for a client, scan the pages that earn them money rather than the homepage, send the public report link, and turn this checklist into tickets split by owner. The client sees the evidence without creating an account, and their developers and writers each get a list they can act on.
- 01
Audit the pages that earn revenue
Run the free audit on the 3 to 5 pages the client most wants quoted: pricing, the main product or service page, and the guide with the most traffic. The homepage is rarely the page an AI engine cites.
- 02
Share the report link, not a screenshot
Every report gets a public link that opens without an account, with each failing check, the evidence behind it, and the fix.
- 03
Split the work by owner
Filter this checklist to developer jobs or content jobs and hand each list over as tickets. Critical checks go first, because they decide whether the client can be cited at all.
- 04
Keep a PDF for the record
Use Print / Save as PDF on this page to attach the ticked checklist to the proposal or the monthly report.
- 05
Re-run after the fixes ship
Audit the same URLs again and send the new link, so the client sees which checks moved from fail to pass.
Run the GEO checklist on your page
The audit reports AI crawler access per user agent, extracts the passages most likely to be quoted, and scores entity and evidence signals.
Frequently asked questions
- What is a generative engine optimization audit?
- A generative engine optimization audit (GEO audit) measures whether an AI answer engine can fetch your page, extract a self-contained answer from it, identify who you are, and trust the claim enough to cite you. It is a retrieval and attribution audit, not a ranking audit.
- How to do a GEO audit?
- To do a GEO audit: (1) verify robots.txt allows AI crawlers like GPTBot, PerplexityBot, and ClaudeBot, (2) check that answers are rendered in server HTML rather than hidden behind client-side JavaScript, (3) format core answers into self-contained 40-60 word definition passages, (4) declare clear entity schema (Organization, Person) with sameAs profile links, and (5) publish a valid llms.txt file. You can run all these checks automatically with IMVASA free audit tool.
- What is a GEO readiness audit?
- A GEO readiness audit evaluates whether your content, schema, and server infrastructure are prepared to be retrieved, synthesized, and cited by generative engines like Perplexity, ChatGPT Search, Gemini, and Google AI Overviews.
- What is included in a GEO site audit?
- A GEO site audit includes 50 checks across 7 groups: AI crawler access in robots.txt, answer readiness and definition density, semantic chunkability, entity clarity and sameAs links, verifiable evidence and citations, and AI-native discovery files like llms.txt.
- What is the difference between an SEO audit and a GEO audit?
- An SEO audit optimises for ranked links in Google. A GEO audit optimises for being retrieved and cited by AI answer engines like ChatGPT, Perplexity, and Google AI Overviews. They overlap on crawlability and structure but diverge sharply on answer readiness, entity clarity, and evidence.
- What is the difference between GEO, AEO and generative search optimization?
- Generative engine optimization (GEO) and generative search optimization are two names for the same work: getting your pages retrieved and cited in AI-generated answers from ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews. Answer engine optimization (AEO) is the older term, from featured snippets and voice assistants, and aims at being the single direct answer. The audit is the same for all three: crawler access, answer-ready passages, entity clarity, structured data and evidence.
- What is a structured data completeness audit?
- A structured data completeness audit checks every JSON-LD and microdata block on a page for the properties its Schema.org type needs: name on Organization, headline on Article, price and priceCurrency on Offer. A missing required property is an error; a missing recommended one, like sameAs, author or brand, is a warning. IMVASA runs it as part of the free audit.
- Does Google penalize AI-generated content?
- Not for being AI-generated. Google judges the page, not the tool: AI-assisted content has to meet the same Search Essentials and spam policies as anything else. What it acts on is scale without value, generating many pages that add nothing for users, which falls under its scaled content abuse policy. Fact-check and review everything before it goes live.
- Can I use this GEO audit checklist for client audits?
- Yes. Run the free audit on the client's key pages, share the public report link, which opens without an account, and use this checklist's developer and content filters to split the fixes into tickets. Print / Save as PDF gives you a copy for the proposal or the monthly report.
- Which checks matter most for AI search visibility?
- In order: AI crawler access in robots.txt, server-rendered content, a direct answer in the opening passage, entity clarity through Organization or Person schema with sameAs links, and sourced statistics. Most sites fail on the first two without knowing it.
- Is there a free tool to audit generative engine optimization?
- Yes. IMVASA Audit Tool runs this GEO checklist alongside a full technical SEO audit on any public URL, free and without signup, and shows the exact passages an AI engine is most likely to quote.
- Does llms.txt actually help AI search visibility?
- It does not affect Google ranking. It gives AI crawlers and agents a curated map of your site, which mainly helps large or documentation-heavy sites. It is cheap to add and it is the file most often missing when a GEO audit runs.
Keep going
SEO audit checklist
Every technical and on-page check, explained.
Free GEO audit tool
Score your AI search visibility on any URL.
Free SEO audit tools compared
Honest look at the free tiers, including Semrush alternatives.
Free SEO audit MCP
Run audits from ChatGPT, Claude or Cursor.
Free SEO tools for every site
robots.txt, sitemap.xml and llms.txt builders.