How to Run a GEO Audit: A Step-by-Step Checklist (2026)

A GEO audit tells you why AI engines do or do not cite you, before you spend a dollar trying to fix it. This is the exact six-step checklist we run at Astral: baseline your AI visibility, check crawlability and bot access, score content extractability, validate schema, assess authority, and set up measurement. By the end you will have a scorecard that turns vague "we should do GEO" into a prioritized roadmap.

What a GEO audit is, and why it comes first

A GEO audit is a structured review of how ready your website is to be discovered, parsed, trusted, and quoted by generative engines: ChatGPT, Perplexity, Google AI Overviews, Gemini, and Claude. It is the diagnostic step that comes before any spend on content or tooling. If you are new to the discipline itself, start with what GEO actually is; this article assumes you know the goal and want to measure where you stand.

The reason to audit first is blunt: you cannot fix what you have not measured. Most teams that "start doing GEO" skip straight to writing FAQ pages or adding schema, then have no way to know if any of it worked. An audit gives you a baseline. Six months later you compare against it and know whether your citation share actually moved, instead of guessing. It also stops you from spending on the wrong layer. Adding schema to a site that blocks GPTBot in robots.txt is wasted effort, and only an audit surfaces that ordering.

Before you start: the tools and baseline setup

The good news is that a complete first-pass GEO audit costs nothing. You can automate parts of it later with paid trackers, but the manual version proves whether you have a problem worth paying to solve. Here is the minimum kit.

• The four engines. Accounts on ChatGPT, Perplexity, Gemini, and Claude. Perplexity and AI Overviews expose inline citations, which makes them easiest to verify.
• A spreadsheet. One tab for your visibility log, one for the scorecard. This is your audit's source of truth.
• Google Rich Results Test and the Schema.org Validator. Both free, for the structured-data step.
• A crawler. Screaming Frog is free up to 500 URLs and surfaces rendering, indexability, and schema gaps at scale.
• GA4. Free, for the measurement step and to catch AI referral traffic you are already getting.
• Your robots.txt and llms.txt. Open yourdomain.com/robots.txt in a browser; you will read it directly in step 2.

If you want to skip the manual visibility logging, a dedicated tracker will run hundreds of prompts on a schedule for you. For a true baseline, though, run at least one manual pass yourself so you understand what the tool is measuring.

Step 1 — Run your AI visibility baseline

This is the heart of the GEO audit and the step most teams get wrong by treating it as a single snapshot. AI answers are non-deterministic, so one prompt on one day tells you almost nothing. You are measuring a distribution, which means repetition.

List your money prompts

Write the 15 to 30 questions a real buyer would ask an AI engine before choosing a product like yours. Not keywords, full natural-language questions. Mix categories:

• Category prompts: "What is the best GEO agency for a B2B SaaS company?"
• Comparison prompts: "Astral vs [competitor] for AI search optimization."
• Problem prompts: "How do I get my site cited by ChatGPT?"
• Branded prompts: "What does Astral do?" and "Is Astral legit?"

Prompt the engines and log everything

Ask each question in all four engines. For every answer, log four things: whether your brand is mentioned, your position or prominence in the answer, which sources got cited, and which competitors appeared. Repeat the full set weekly for at least two to three weeks so randomness averages out. Your log should look like this:

Two numbers fall out of this log. Your citation share is the percentage of prompts where you are cited at all. Your share of voice weights that by prominence and compares it to competitors. Those two numbers are your baseline. For the full set of metrics and how to report them over time, see our guide to tracking and measuring GEO performance.

Step 2 — Audit crawlability and AI bot access

You can write the most quotable content on the internet, but if the AI crawlers cannot fetch it, none of it matters. This step is the cheapest, highest-leverage part of the GEO audit because the fixes are usually a few lines of configuration.

Check which AI bots you allow

Open your robots.txt and look for explicit allow or disallow rules for the AI user agents. Many sites silently block them, often because a security plugin or CDN rule did it by default. The agents that matter in 2026:

Confirm none of these are disallowed. A minimal permissive block looks like this:

User-agent: GPTBot
Allow: /

User-agent: PerplexityBot
Allow: /

User-agent: ClaudeBot
Allow: /

User-agent: Google-Extended
Allow: /

Check llms.txt, rendering, and indexability

• llms.txt. Check whether yourdomain.com/llms.txt exists. This emerging file gives engines a curated map of your most important content. If you do not have one, that is a flagged gap for your roadmap.
• JavaScript rendering. Many AI crawlers do not execute JavaScript as fully as Googlebot. View your key pages with JS disabled, or use your crawler's rendered-vs-raw comparison. If your main content only appears after client-side rendering, AI engines may see a blank page. Server-side render or pre-render critical content.
• Indexability. Confirm important pages are not accidentally noindex, blocked in robots.txt, or buried behind broken canonicals. A crawler surfaces all three in one pass.

Step 3 — Score content extractability

Once engines can reach your pages, the question becomes whether they can lift a clean, self-contained answer from them. Classic SEO rewards comprehensive long pages; GEO rewards passages an engine can quote without surrounding context. Audit a representative sample of your key pages against these criteria.

• Answer-first structure. Does each section lead with a direct answer in the first one or two sentences, then elaborate? Engines lift the lead, not the buildup.
• Self-contained chunks. Can a single paragraph stand alone and still make sense if quoted in isolation? Avoid "as mentioned above" dependencies.
• Question-shaped headings. Do your H2s and H3s match how people actually ask, rather than clever marketing labels? "How much does GEO cost?" beats "Investment tiers."
• Freshness signals. Are dates visible and current? Engines favor recently updated, dated content for anything time-sensitive.
• Clarity and specificity. Concrete numbers, named entities, and plain claims get quoted; vague marketing language gets skipped.

Score each page pass or fail on these five, and you will quickly see the pattern. Most sites fail on answer-first structure and self-contained chunks because they were written to be read top to bottom, not sampled. Our deep dive on writing content that gets cited by AI is the companion fix for whatever this step exposes.

Step 4 — Audit your structured data

Schema markup is not magic, but it removes ambiguity about your entities and claims, which makes you easier to parse and cite. The audit here is simple: list what you have, compare it to what each page type needs, and validate it.

Run your key pages through the free Google Rich Results Test and the Schema.org Validator, and fix any errors or warnings. Then check that the schema actually matches the visible content; mismatched or stale markup hurts more than missing markup. For which types move the needle and how to implement them cleanly, see our practical guide to schema markup for GEO. The tooling is trivial; choosing the right types and keeping them accurate is the real work.

Step 5 — Assess authority and entity signals

AI engines decide whom to trust using many of the same signals classic search uses, plus a heavier reliance on third-party corroboration. This step is the slowest to improve, so audit it honestly and plan a long runway.

• Backlinks and referring domains. Do authoritative, relevant sites link to you? Engines lean on the same authority graph that powers ranking.
• Brand mentions. Are you discussed on sites the engines already trust, even without a link? Unlinked mentions still build your entity.
• Entity and NAP consistency. Is your name, description, and key facts identical across your site, social profiles, and directories? Conflicting facts make engines uncertain, and uncertainty means they skip you.
• Third-party citations and reviews. Do you appear on the review sites, roundups, and "best of" lists engines pull from? In many categories, being on a G2 or comparison list is worth more than another blog post.
• Knowledge-graph presence. Does a Wikipedia, Wikidata, or Google Knowledge Panel entry exist for your brand? These are strong entity anchors engines defer to.

The fastest-improving sites in AI search are rarely the ones publishing the most. They are the ones whose facts are consistent everywhere an engine looks, so the model never has a reason to doubt them.

Score this area on a simple low / medium / high. If your entity signals are fragmented, that is often the real ceiling on your citation share, and no amount of on-page work overcomes it. Authority and consistency are the slowest area to move, so if this step is where you stall, treat it as a multi-quarter program rather than a quick fix.

Step 6 — Set up GEO measurement

The final audit step is making sure you can keep measuring after the audit ends. Without this, your baseline rots and you are guessing again within a quarter.

• GA4 AI-referral segment. Build a custom segment or channel grouping for referrals from chatgpt.com, perplexity.ai, gemini.google.com, and Copilot. This captures the clicks AI is already sending you, even though much AI value is zero-click.
• Citation tracking cadence. Schedule your visibility log from step 1 to repeat weekly. Put it on a calendar or hand it to a tracker. Consistency beats precision here; the same prompts every week reveal the trend.
• A reporting baseline. Record your starting citation share and share of voice so the next audit has something to compare against.

Attribution in AI search will never be perfect, because so much of the value is brand exposure that never produces a click. Track the trend, not a single number, and pair referral data with your citation log for the honest picture.

Turning the GEO audit into a roadmap

An audit that ends in a pile of notes changes nothing. The point of the GEO audit is to produce a prioritized roadmap, and the cleanest way to do that is a scorecard. Score each of the six areas from 1 to 5, weight them, and the lowest weighted scores become your first projects.

The sequencing almost always follows the same logic regardless of scores. Fix bot access first because it is binary and caps your ceiling. Knock out schema next because it is fast. Then commit to the two slow, high-weight areas, content extractability and authority, as ongoing programs rather than one-time tasks. Re-run the whole audit next quarter and compare. If you run a SaaS specifically, our GEO playbook for SaaS maps this roadmap onto product-led growth, and if your priority is Google's surface, ranking in Google AI Overviews goes deeper on that engine.

Common GEO audit mistakes to avoid

A few traps recur often enough to call out, because each one quietly invalidates the audit.

• Sampling once. One prompt on one day is noise. AI answers vary run to run; you need repetition to find the real distribution.
• Auditing content before checking access. If bots are blocked, your content scores are irrelevant. Always do step 2 before you agonize over step 3.
• Treating schema as a ranking hack. Schema clarifies; it does not force citations. Markup that contradicts your visible content backfires.
• Ignoring competitors. Your citation share only means something relative to who is winning the prompts you want. Always log who shows up instead of you.
• Auditing once and stopping. The single biggest mistake. A GEO audit is a quarterly baseline, not a project you close.

Do this well and the audit pays for itself immediately: it stops you from spending on the wrong layer and gives you a number to beat. The teams winning in AI search are not the ones doing the most; they are the ones who measured, found their real gap, and fixed it on purpose.

Frequently asked questions

What is a GEO audit?

A GEO audit is a structured review of how ready your site is to be found and cited by AI search engines like ChatGPT, Perplexity, Gemini, and Google AI Overviews. It baselines your current AI visibility, then checks the six things that determine it: crawlability and bot access, content extractability, structured data, authority and entity signals, and measurement. The output is a scorecard of gaps you can prioritize and fix.

How long does a GEO audit take?

A focused first-pass GEO audit takes about one to two days of work for a single site, plus a week or two of passive monitoring to build a reliable visibility baseline. The visibility baseline is the slow part because you need to prompt the engines repeatedly to separate signal from the randomness in AI answers. The technical and schema checks can usually be done in a single afternoon with free tools.

Can I do a GEO audit for free?

Yes. The entire audit in this checklist can be run with free tools: the AI engines themselves for visibility, a browser and your robots.txt for bot access, the Google Rich Results Test and Schema.org Validator for structured data, and GA4 for measurement. Paid trackers automate the visibility baseline across hundreds of prompts, but you can prove whether you have a problem worth solving for zero dollars first.

How often should I run a GEO audit?

Run a full GEO audit once per quarter, and track your visibility baseline continuously between audits, ideally weekly. AI engines, their models, and their citation behavior change fast, so a one-time audit goes stale within months. The quarterly cadence catches structural drift, while weekly visibility logging catches sudden swings in which sources the engines favor.

What is the difference between a GEO audit and an SEO audit?

An SEO audit checks whether you can rank in a list of blue links; a GEO audit checks whether you get cited inside a generated answer. They overlap heavily on crawlability, site structure, and authority, but a GEO audit adds AI bot access, answer-first content extractability, entity consistency, and visibility tracking across AI engines rather than keyword rankings. Most teams should run both because AI engines still lean on classic search signals.

Should I run my own GEO audit or hire an agency?

You can run the baseline audit in this checklist yourself with free tools, and you should, because it tells you whether you even have a problem. Hire an agency when the audit reveals gaps you lack the time or expertise to fix, such as rebuilding content for extractability across hundreds of pages or repairing fragmented entity signals. A good agency starts with this same audit and then executes the roadmap it produces.