AI Visibility Audit Checklist
Ten checks that show whether AI crawlers can reach your site, whether your pages can be quoted, and whether buyers hear your name when they ask an engine.
An AI visibility audit checks three layers in order: can each engine's crawler reach your pages, can it read and quote them once fetched, and are you actually named when a buyer asks a real question. Below is the checklist we run on our own site and every client's before anything else.
This checklist is the practical companion to AI brand visibility: the measurement discipline that guide describes starts here, with a one-time pass to find out where you actually stand.
On this page
What are the three layers, and why in this order?
Because each one is a precondition for the next. A crawler that cannot reach your page cannot read it. A page that cannot be read cannot be quoted. And a page that is never quoted cannot be the reason you get named — so fixing presence before fixing access is fixing the wrong end of the chain.
Most teams want to start at layer three, because that is the layer that actually shows up in a chat window and feels like the real problem. Resist that. We have opened audits where a company spent months on presence work — pitching listicles, cleaning up entity data — while a firewall rule silently blocked two of the six crawlers they were trying to influence the entire time. The order in this checklist is not a stylistic choice; it is the order that avoids wasting the presence work on a site that was never fetchable in the first place.
| Layer | What it checks | Typical failure |
|---|---|---|
| Access | Can each engine's crawler fetch the page at all | A firewall or bot-protection rule blocking the user agent, invisible in the robots.txt file |
| Readability | Is there anything a crawler can actually quote once it fetches the page | Content that only renders after JavaScript runs, or sits behind a form |
| Presence | Does a real buyer question actually surface your name | Third-party pages, not your own site, are what gets cited instead |
1–4: Is your site reachable by every crawler that matters?
1. List every crawler you should allow
Check each engine's published list by name rather than trusting a wildcard rule — the names have changed before, and a stale list silently misses a new one [1].
2. Fetch your homepage as each crawler
Use curl with each crawler's real user agent string against your live site, not staging, and confirm a 200 and full HTML.
3. Check the edge, not just robots.txt
Hosting providers and CDNs ship bot-protection presets that block AI user agents by default, with no error and no visible traffic drop.
4. Confirm you are not suppressing snippets
A directive that forbids quoting forbids the exact thing you want an assistant to do with your page.
5–6: Once fetched, is there anything worth quoting?
5. Compare raw HTML with the rendered page
Fetch the page without JavaScript and read what is actually there. Content that only exists after a script runs may not exist at all to a retrieval crawler.
6. Check that the first sentences of each section stand alone
Read a section aloud starting at the first sentence and stop after three. If it is a complete, correct answer, it is quotable; if you were still setting up context, it is not.
7–8: Are you named when a buyer actually asks?
7. Build a real prompt set
Twelve to forty questions pulled from sales calls and support tickets, covering category questions, comparison questions naming competitors, and questions about you by name.
8. Run it weekly, on every engine, for at least three weeks
A single run is noise. The same prompt asked two days apart returns a meaningfully different set of sources, so judge a trend, not a snapshot [3].
This is the step where most audits find their real gap. A site can pass every access and readability check and still be almost invisible in answers, because the citations an engine is actually using come from third-party pages, not the brand's own site.
Run the same prompt set on every engine you care about, not just the one you personally use most. The four major assistants disagree with each other more than most teams expect — one tracked sample found all four models agreeing on the same cited domain for the same question only 1.7 percent of the time [3]. A company that only checks ChatGPT because that is what the marketing team uses day to day can be entirely invisible on Perplexity or Gemini and never find out from a single-engine habit.
Log the result in a plain spreadsheet: prompt, engine, date, named (yes/no), cited (yes/no), and which URL got cited if not yours. This is unglamorous and it is the entire method. Every tool that automates this later is automating exactly this spreadsheet, so building it once by hand is worth doing before paying for the automated version.
9: Is your entity data the same everywhere someone would check it?
Search your own company name and check what comes back: your site, your social profiles, your review and directory listings. Look specifically for a mismatched legal name, an old address, a discontinued product still listed, or a founder who has left. Inconsistent entity data is a quiet source of both weak presence and outright wrong answers, because it gives an assistant conflicting signal to resolve instead of one clear fact to repeat.
Give this check a fixed list rather than doing it from memory: legal name, short name, founding year if you publish one, headquarters city, founder names, and your core one-sentence description of what you do. Pull up your site, your LinkedIn page, any review platform profile, and any directory listing side by side and compare each field. A ten-minute pass with this list catches more than an open-ended "does everything look right" scan, because it forces you to check the specific fields an assistant is most likely to repeat.
10: Are you covered on the third-party pages your engines already cite?
Look at the sources your presence testing turned up, specifically the ones cited instead of you. That list is your outreach priority, not a generic directory search. Get featured in best-of lists covers the outreach mechanics once you have this list in hand, and it is worth doing before you re-run the audit, because this is usually where the biggest single jump in presence comes from.
What do you do with a finding that isn't a gap, but a wrong answer?
Treat it separately and urgently. A gap means you are absent; a wrong answer means the assistant is actively repeating something false about you, and that is worth fixing before anything else on this list. Fix wrong AI answers about your brand walks through the correction path for both your own site and the third-party pages an assistant might be reading from.
How do you turn ten checklist items into one report?
Score each item pass, partial, or fail, note the specific fix, and give it an owner and a rough date — a checklist with no owner attached to each finding is a list nobody acts on. For the presence layer specifically, convert your weekly runs into the handful of numbers that are worth tracking over time rather than re-explaining the raw log every month; AI visibility metrics covers which numbers are worth keeping and which ones are noise dressed up as a KPI.
Re-run the whole checklist on a schedule, not just once. Access and readability rarely regress on their own, but they do regress when a security team rotates a firewall vendor or a developer ships a new bot-protection rule without checking it against your crawler allowlist. Presence and entity data move for reasons entirely outside your control — a competitor gets included in a listicle, an engine changes its retrieval behavior, a directory listing you never touched goes stale. A quarterly pass through all ten items catches the kind of drift that a team focused only on new content will otherwise miss for months.
Our own AI visibility audit runs this exact checklist for you, across every engine in scope, if you would rather hand it off than run it by hand each quarter.
Questions people also ask
How long does a full AI visibility audit take?
Doing it yourself with the checklist below takes a focused day or two for the access and readability checks, plus a week of runs for the presence testing, since a single prompt run is not enough to trust. A paid audit that covers several engines and languages usually runs one to two weeks.
Do I need special tools to run this checklist?
No. Crawler access can be tested with curl and a user agent string, readability with your browser's view-source, and presence testing with the chat interface of each engine. Tools help once you are running this weekly across many prompts, not before.
Which item on this checklist do most companies fail first?
Presence. Access and readability are usually fine on a modern site with no unusual firewall rules. The gap almost always shows up when you actually run buyer questions and see who gets named instead of you.
Should I run this checklist once or on a schedule?
Once to find the current state, then on a schedule after that. Access and readability rarely change month to month once fixed; presence and entity data are worth checking monthly, because third-party pages and model behavior both shift.
Sources
- Robots.txt introduction and guide — Google Search Central (checked 23 September 2026)
- AI features and your website — Google Search Central (checked 23 September 2026)
- I tracked 13,184 AI citations across 4 LLMs: here's what I learned — Orbit Media Studios (checked 23 September 2026)
Get this checklist run for you across every engine you care about
Book a free 30-minute consultation with Ibrahem and Jeremy. We look at your site while we talk, and you leave with a prioritised plan.
Who wrote this
Jeremy ZavaCo-founder
Leads growth and client partnerships.
Works on: CRO, SEO
Reviewed before publishing by Ibrahem Helal, Co-founder.
- How to Get Featured in Best-Of ListsAI assistants cite third-party listicles more than brand marketing pages. Here's how to find the lists that already rank and get added to them.
- How to Fix Wrong AI Answers About Your BrandWhen an assistant gets a fact about your company wrong, here's how the correction actually reaches the model, and what you can fix directly today.
- AI Visibility Metrics That Actually MatterCitation rate, mention rate and share of voice measure different things. Here are the AI visibility metrics worth tracking, and how to read them.
- AI brand visibility: how to measure and grow itHow often ChatGPT, Perplexity, Gemini and Google AI Mode name or cite your brand, how to measure it on a schedule, what moves the number and what it costs.