TL;DR: A GEO audit answers three questions in order: can AI engines reach your content, do they cite you today on the queries that matter, and what specifically blocks you where they do not. Run it in five steps: crawl access, index coverage, citation baseline, page-level content grading, and entity check. Output is a ranked fix list, not a score.
Most companies start GEO by publishing new content, which is like renovating a store before checking whether the doors open. An audit runs the checks in dependency order, because a schema fix means nothing if GPTBot is blocked at robots.txt, and a content rewrite means nothing if the real gap is that engines do not recognize your brand as an entity. Here is the framework we run, step by step, with the pass and fail criteria for each.
Step 1: Can AI crawlers access your site?
Check robots.txt, your CDN, and your rendering before anything else, because blocked crawlers zero out every other effort. Verify that GPTBot, OAI-SearchBot, ClaudeBot, PerplexityBot, Google-Extended, and Bingbot are not disallowed in robots.txt, and that your priority sections (blog, guides, service pages, comparisons) are crawlable.
robots.txt is not the whole story. CDN and bot-protection layers like Cloudflare frequently challenge or block AI user agents silently; pull server logs and confirm each bot actually receives 200 responses, not 403s or challenge pages. Then check rendering: most AI crawlers do not execute JavaScript, so content that only exists client-side is invisible to them regardless of permissions. Fetch key pages with JavaScript disabled and confirm the answer content is in the raw HTML.
The blocking decision itself deserves thought rather than defaults, and the tradeoffs differ per bot; we walked through them in should you block AI crawlers. Pass criteria for this step: every retrieval bot you want citations from returns 200s on your money pages, and the content is server-rendered.
Step 2: Are you indexed where the engines shop?
Confirm coverage in both Google and Bing, because each index gates different engines. Google’s index feeds AI Overviews, AI Mode, and Gemini grounding. Bing’s index feeds ChatGPT search and Copilot, and it is the one most teams have never checked. Run site: queries and check Bing Webmaster Tools coverage reports the same way you check Google Search Console.
Bing gaps are the most common single finding in our audits. A site can rank well on Google while half its blog is absent from Bing, which silently caps ChatGPT visibility at zero for those pages. If you find gaps, submission and IndexNow get you covered within days, and the fix is essentially free. Pass criteria: your citable pages appear in both indexes, and new content gets indexed in both within a week of publishing.
Step 3: What is your citation baseline?
Build a query set of 20 to 50 real buyer prompts, run them across every major engine, and record who gets cited, including competitors. This is the heart of the audit. Pull the prompts from how buyers actually ask: your sales call questions, People Also Ask boxes, autocomplete, support tickets. Then run each identical prompt through ChatGPT search, Perplexity, Google AI Mode or AI Overviews, Gemini, and Copilot, and log every citation in a sheet: engine, query, position, domain cited.
Two numbers fall out of this. Your citation rate: the share of prompt-engine combinations where you appear. Your share of voice: your citations divided by total citations in your category. Record competitor numbers with the same discipline, because the gap between you and the most-cited competitor is the actual size of the opportunity. Tools like Otterly, Profound, and Peec automate the re-running at scale, and we compared the tooling options in how to track your AI search visibility.
The baseline also classifies your gaps. Cited on Perplexity but absent on ChatGPT for the same query usually means a Bing index or freshness issue. Competitors cited consistently while you never appear anywhere usually means an entity gap, which step five diagnoses. Same-query inconsistency across engines is normal; per-engine patterns are the signal.
Step 4: Do your pages deserve citations?
Grade each priority page against the signals engines quote: direct answers, statistics, quotable sentences, structure, and freshness. Research on citation behavior consistently favors pages with specific numbers, named sources, definitive phrasing, and recent dates. Score every priority page on a simple rubric, one point each:
- The target question is answered in the first two to three sentences.
- Subheadings are phrased as questions or claims, answered in their first 40 words.
- The page contains at least three specific, sourced statistics or facts.
- There is FAQ, Article, or relevant schema markup that parses without errors.
- The page shows a visible, truthful updated date within the last 12 months.
- An identifiable author with a linked bio stands behind it.
Pages scoring 5 to 6 are citation-ready; 3 to 4 need targeted edits; 0 to 2 need rewrites or retirement. This turns a vague content audit into a ranked work queue, and the scoring rubric doubles as your writing standard going forward, the same one behind our 30-step GEO checklist.
Step 5: Do engines know who you are?
Ask the engines about your brand directly and audit what comes back, because entity confusion silently suppresses citations everywhere else. Prompt each engine with “what is [company],” “is [company] legitimate,” and “[company] vs [competitor].” Wrong facts, outdated descriptions, or confusion with a similarly named business all indicate weak entity signals.
Then verify the corroboration layer: consistent name, address, and description across your site, LinkedIn, Crunchbase, industry directories, and review platforms; sameAs schema linking your profiles; and third-party mentions in press or on platforms engines cite. Brand web mentions correlate with AI citation rates far more strongly than backlinks do, roughly 0.664 versus 0.218 in published correlation studies, which is why a technically clean site with no off-site footprint still fails the citation test.
What does the audit output look like?
A ranked fix list ordered by dependency and effort, plus the baseline sheet you will re-run monthly. Resist the urge to compress the audit into a single score. A score hides the dependency logic that makes the audit useful: a site can score 80 out of 100 and still earn zero citations because the missing 20 points are all in step one. The fix list format keeps causality visible, and it gives whoever executes the work an order of operations instead of a grade to argue about. A typical output: unblock two bots at the CDN (day one), submit 40 missing URLs to Bing (week one), add direct-answer openings and FAQ schema to the twelve pages scoring 3 to 4 (weeks two to four), fix the wrong founding date engines repeat (week two), then write the three genuinely missing pages the query set exposed (month two). The re-run cadence is monthly: same prompts, same engines, same sheet, so that movement is measurable rather than anecdotal.
FAQ
How long does a GEO audit take? Two to five working days for a typical small or mid-size site: half a day for crawl and index checks, one to two days for the citation baseline, one to two days for page grading and the entity check. Enterprise sites take longer at the page-grading step.
How often should you repeat it? Re-run the citation baseline monthly and the full audit quarterly. Engines change retrieval behavior often enough that a six-month-old audit is a historical document.
What tools do you need? Minimum viable stack: robots.txt and log access, Google Search Console, Bing Webmaster Tools, a spreadsheet, and manual prompt runs. Citation-tracking tools like Otterly, Profound, or Peec replace the manual re-running once you are operating at scale.
Can you do a GEO audit without a developer? Steps two through five, yes. Step one usually needs whoever controls the CDN and robots.txt to confirm and fix bot access, which is often a 30-minute change once identified.
How is a GEO audit different from an SEO audit? An SEO audit optimizes for ranking positions in one index. A GEO audit spans multiple engines with different indexes and retrieval habits, measures citations rather than rankings, and adds checks SEO ignores: AI bot access, JavaScript visibility to non-rendering crawlers, quotability grading, and entity accuracy inside model answers.
The bottom line
A GEO audit replaces guessing with a dependency-ordered fix list: access, index, baseline, content, entity. Run it once and you know exactly why you are absent from the answers your buyers read; re-run the baseline monthly and you know whether the fixes are working. If you would rather see your baseline than build it, contact us and we will run the full prompt set for your category, or start with the ROI calculator.
Tagged