Your website ranks on page one of Google, yet ChatGPT recommends a competitor when a buyer asks for the best option in your category. That gap is exactly what a GEO audit is built to close. As AI assistants like ChatGPT, Perplexity, Google AI Overviews, and Claude become the first stop for buying decisions, classic search rankings are no longer enough. You also need to be readable, quotable, and trusted by the systems that generate those answers.
A GEO audit (Generative Engine Optimization audit) is a structured review of how visible your site is to AI search engines and how likely those engines are to cite you. It checks whether AI crawlers can reach your content, whether that content is structured for extraction, and whether your brand carries the authority and entity signals that make models confident enough to name you. This guide walks through a complete AI visibility audit across the six GEO pillars, with what to check, how to check it, and how to fix each issue.
Traditional SEO optimizes for a ranked list of blue links. GEO optimizes for a synthesized answer where usually one to three sources get named and the rest disappear. If your pages are sound for Google but invisible to GPTBot or PerplexityBot, you hand that answer space to competitors. A GEO audit tells you where you stand before you spend on content or PR.
The fastest way to get a baseline is to run an automated scan. Our free AI Visibility Checker scores your domain across the same six GEO pillars covered below and shows how you compare against competitors in minutes. Use it to get numbers first, then work through the manual checks in this guide to understand and fix what the scan surfaces. For the strategic context behind all of this, start with our pillar guide to generative engine optimization.
If AI crawlers cannot fetch your pages, nothing else in this audit matters. This is the single most common and most costly GEO failure, and it is usually caused by a stale robots.txt file or an over-aggressive firewall rule.
Confirm that the major AI user agents are not blocked: GPTBot and OAI-SearchBot (OpenAI), PerplexityBot (Perplexity), Google-Extended (Gemini and AI Overviews training), ClaudeBot (Anthropic), and Bingbot (which feeds Copilot).
Open yourdomain.com/robots.txt in a browser and look for any Disallow: / rules tied to these agents. Then check server-level blocks: many WAF and CDN providers (Cloudflare in particular) now ship default rules that block AI bots. Review your CDN bot-management settings and scan raw server logs for these user agents to confirm they are actually crawling and returning 200 status codes, not 403s.
Explicitly allow the agents you want. A permissive baseline looks like this:
User-agent: GPTBot
Allow: /
User-agent: OAI-SearchBot
Allow: /
User-agent: PerplexityBot
Allow: /
User-agent: Google-Extended
Allow: /
User-agent: ClaudeBot
Allow: /Then whitelist these agents in your CDN or WAF. Decide deliberately: blocking Google-Extended stops Gemini training but does not affect Google Search indexing, while blocking GPTBot removes you from ChatGPT browsing citations. For most brands chasing AI visibility, allow all of them.
The llms.txt file is an emerging standard that gives AI models a clean, curated map of your most important content in Markdown. Think of it as a sitemap written for language models rather than search crawlers, pointing them to your definitive pages without the noise of navigation, ads, and scripts.
Check whether a file exists at yourdomain.com/llms.txt, whether it lists your priority pages, and whether the links resolve to live, high-value URLs.
Visit the URL directly. If it returns a 404, create one. Keep it simple: an H1 with your brand name, a short blockquote summary of what you do, and grouped Markdown links to your core product, service, and cornerstone content pages. This is a low-effort, low-risk addition. Adoption by AI engines is still early and not universal, so treat it as a helpful signal rather than a guaranteed ranking factor, and do not expect it to compensate for weak content elsewhere.
Schema markup translates your content into a machine-readable format that removes ambiguity. When a model can read explicit Organization, Author, Product, or FAQPage data, it spends less effort guessing what your page means and more confidence citing it.
Look for the schema types that matter most for AI answers: Organization (with sameAs links to your social and knowledge-base profiles), Article with a named author, FAQPage for question-and-answer content, Product and Review for commercial pages, and BreadcrumbList for context.
Run key URLs through Google’s Rich Results Test and the Schema.org validator. Confirm the markup is present, valid, and accurate rather than boilerplate copied across every page. Pay special attention to whether your Organization schema consistently declares the same name, logo, and sameAs profiles sitewide.
Add JSON-LD in the page head for any missing types. On WordPress, a plugin like Yoast or Rank Math handles most of this, but verify the output rather than trusting defaults. The highest-leverage fix is a strong, consistent Organization block on every page, since it anchors your brand as a recognizable entity that models can connect across the web.
AI engines favor content they can lift cleanly into an answer. A brilliant insight buried in a wall of text is far less citable than a direct, self-contained statement. This pillar is where good SEO writing and good GEO writing diverge most.
Audit your top pages for: clear question-style H2s and H3s, direct answers placed in the first sentence under each heading, short scannable paragraphs, definition sentences that stand alone, and structured elements like lists, tables, and step blocks. Also confirm the content renders in raw HTML, not only after JavaScript execution, since many AI crawlers do not run JS.
View the page source (Ctrl+U) or fetch the raw HTML with a tool like curl and confirm your main content is present without a browser rendering it. Then read your headings as standalone questions and ask whether each opening sentence answers the question on its own. A simple test: copy any single paragraph out of context and see if it still makes a complete, quotable point.
Rewrite headings as the questions your audience actually asks, then answer them immediately before adding nuance. Break long paragraphs into two or three sentences. Convert comparisons into tables and processes into numbered steps. Move critical content out of JavaScript-only components and into server-rendered HTML. For a deeper playbook on making pages quotable, see our guide on how to get cited by AI.
Models are tuned to prefer sources that demonstrate experience, expertise, authoritativeness, and trust. They also need to recognize your brand as a distinct entity, not just a string of words. Weak entity signals are why some well-written sites still never get named in AI answers.
Verify that articles have real, credentialed authors with bio pages, that your brand has a consistent presence across third-party sources (Wikipedia, Wikidata, G2, industry directories, reputable press), and that facts are backed by citations, original data, or first-hand examples rather than generic filler.
Ask an AI assistant directly: “What is [your brand] and what do they do?” The accuracy and confidence of the answer reveals how well models understand your entity. Then search for your brand across the reference sites above and note where you are absent or described inconsistently.
Add detailed author bios with credentials and link them via Person schema. Build or correct your Wikidata entry, keep your name and description consistent everywhere, and pursue mentions on trusted third-party sites. Publish original research, case studies, and named examples that only your team could produce, since first-hand experience is the hardest signal for competitors to copy.
The classic technical basics still apply, because a page that loads slowly, breaks on mobile, or returns errors is harder for any crawler to process reliably.
Confirm fast load times and healthy Core Web Vitals (PageSpeed Insights), a valid and submitted XML sitemap, clean internal linking so important pages are reachable, HTTPS across the site, no broken links or redirect chains, and a mobile-friendly layout. These are table stakes: they will not win you AI citations on their own, but any failure here quietly undermines all five pillars above.
Prioritize crawl reliability: fix 404s and redirect loops, compress images, cache aggressively, and keep the sitemap current. Ensure every priority page is no more than a few clicks from the homepage and linked from related content.
Work through this checklist in order. Each item maps to one of the six pillars above.
GPTBot, PerplexityBot, Google-Extended, and ClaudeBot are allowed in robots.txt and at the CDN or WAF level.llms.txt file exists and links to your priority pages.Organization schema with sameAs is present and consistent sitewide.Article and Author schema; FAQ sections use FAQPage.Not every issue deserves equal urgency. Score each pillar from 0 to 100 based on the checklist items you pass, then weight your fixes by impact.
llms.txt and remaining technical polish, useful but lower priority given current adoption.A practical rule: a domain scoring below 50 almost always has a crawler or extractability problem to solve before anything else. Sites in the 50 to 80 range usually win the most by strengthening authority and entity signals.
Skip the manual guesswork for your baseline. The TopOnSeek AI Visibility Checker scores your site across all six GEO pillars and benchmarks you against competitors in minutes, so you know exactly where to focus.
Run a full audit quarterly, since AI crawler policies, schema standards, and model behavior change fast. Check crawler access and server logs monthly, because a single CDN update or robots.txt edit can silently block AI bots overnight.
Yes, though they overlap. An SEO audit optimizes for rankings in a list of links. A GEO audit optimizes for being read, trusted, and cited inside an AI-generated answer, which adds new checks for AI crawler access, content extractability, and entity signals that traditional SEO audits ignore.
At minimum: a browser to inspect robots.txt and page source, a schema validator, PageSpeed Insights, and access to your server logs. An automated scanner like our AI Visibility Checker speeds up the baseline, and you can compare specialized options in our roundup of the best GEO tools.
Crawler-access fixes can surface in AI answers within days to a few weeks as bots recrawl. Authority and entity improvements take longer, often one to three months, because models need to encounter consistent signals across multiple sources before they update how they describe your brand.
A GEO audit turns “why isn’t AI recommending us?” into a concrete, prioritized action list. Work through the six pillars in order, clear the crawler and extractability blockers first, then build the authority and entity signals that earn citations. Start with an automated baseline from the AI Visibility Checker, fix what it surfaces using the manual checks above, and re-audit each quarter. The brands that treat AI visibility as a measurable discipline today are the ones models will name tomorrow.
TOS collaborates & develops alongside reputable industry-leading partners
