How to Audit a Competitor's AI Visibility (and Find the Pages Doing the Work)
Most competitor audits stop at a spreadsheet of who got mentioned, which tells you a rival is winning without telling you why. Only 2.9% of AI citations point at a brand's own domain, so the page carrying your competitor almost always belongs to someone else. Here is the seven-step audit that finds it, and the part no audit can fix.
To audit a competitor's AI visibility, run a fixed set of unbranded buyer prompts on each engine several times, record which companies get named and every source the answer cited, then work out which of those cited pages are doing the work. The last step is the audit. The rest is bookkeeping.
Most guides to this stop at a spreadsheet of who got mentioned. That tells you a competitor is winning without telling you why, which makes it impossible to act on. The reason to keep going is in the measurement: only 2.9% of AI citations point at a brand's own domain, while 75.2% point at corporate and competitor pages. A competitor is almost never beating you with their website. They are being carried by somebody else's page, and that page has a URL you can go find.
One honest limit before you start: part of any competitor's advantage will have no page behind it at all, and no audit can hand you an action for that portion.
Two Things That Will Otherwise Waste the Audit
Your competitor list is not a fact you look up. It is an output of how hard you sampled. Holding a set of brands fixed and only increasing the number of responses examined, the roster of companies AI names alongside them climbs from 33 to 282 without settling.
Your Competitor List Is a Function of How Hard You Looked
The same brands at every point. Only the sampling depth changes, and the roster climbs from 33 to 282 companies without flattening.
Source: Ranqo, within-brand rarefaction across the 14 brands with 400 or more competitor-carrying responses. Depth N uses each brand's first N responses by record id, so the subsample is deterministic.
The practical consequence is that "how many competitors do we have in AI answers" has no stable answer, so do not build the audit around that count. Build it around a fixed prompt set and a fixed depth, and compare like with like over time.
One run is not a measurement. Whether a brand is mentioned flips 6.8% of the time between consecutive identical measurements. Audit from a single pass and you will confidently report a competitor gap that is really a coin landing the same way twice.
The Seven Steps
| # | Do this | You end up with |
|---|---|---|
| 01 | Fix a prompt set of 10 to 20 unbranded buyer questions. No company names in them, including yours. | A prompt list you will not change for the length of the audit. |
| 02 | Name the competitors you already believe you have, then treat that list as a hypothesis rather than the answer. | A starting roster, held loosely. |
| 03 | Run every prompt on each engine separately, logged out, at least five times across different days. | A response set large enough that one unlucky answer cannot move it. |
| 04 | For each answer record three things: which companies were named, in what order, and every source cited. | The raw sheet. The third column is the one most audits skip. |
| 05 | Count how often each competitor appears, and note the ones you had never listed. | A ranked appearance table plus a discovery list. |
| 06 | Group the cited sources by domain and by who publishes them, separating third-party pages from the competitor's own site. | The citation trail: the pages doing the work. |
| 07 | Pick the third-party pages that cite a competitor and not you, and turn them into an outreach target list. | A finite, ranked list of pages to go earn a place on. |
Steps one to five are what every guide covers. Six and seven are the audit.
Step 1: Prompts That Can Actually Name a Competitor
The prompt set decides everything downstream, and there are two ways to get it wrong. Naming any company in the prompt, including your own, turns the exercise into a recall test: the engine will discuss whoever you handed it. And asking questions whose answer is a definition or a yes-or-no gives you a response with no vendors in it at all, which is a wasted run.
Write the questions a buyer types before they know who to buy from. Ten to twenty is plenty. Keep them short; real buyer questions are much shorter than the ones marketers write for them. Then freeze the list, because changing it later restarts the series.
Step 4: What to Record, Including the Column Everyone Skips
| Column | Why it earns its place |
|---|---|
| Prompt, engine, date, run number | Without these four you cannot tell a real change from run-to-run noise later. |
| Competitors named | The appearance count. This is the number most audits stop at. |
| Order of mention | Position moves before presence does, so it is the earlier warning signal. |
| Every URL cited | The actual mechanism. A competitor wins because specific pages exist, and those pages are findable. |
| Does the cited page mention you | Splits the target list into pages to get added to and pages to get written into. |
If you record only one thing beyond the competitor names, make it every cited URL. That column is what separates an audit from a scoreboard.
Step 6: Reading the Citation Trail
Group every URL you collected by domain, then split them by who publishes the page: the competitor, or somebody else. Given that third-party pages take the overwhelming majority of citations, most of your list will land in the second pile, and that pile is the actionable one.
| What you see | What it means | What to do |
|---|---|---|
| A ranked list naming them and not you | The highest-frequency pattern. Ranked lists are the single largest content-citation format. | Pitch the editor a ready-to-paste entry. This is the highest-leverage single action in the whole audit. |
| Their own domain cited repeatedly | Rarer than it looks. Own domains take a small minority of citations overall, so a competitor winning this way is winning narrowly. | Look at what the page answers, not how it is marked up. Usually it answers a specific task completely. |
| Editorial coverage you have none of | A public-relations gap wearing a content costume. | Treat as a PR target, not a blog brief. |
| No citation at all behind the recommendation | Common. Engines recommend brands without citing a source for them more often than not. | Nothing to pitch. Note it and move on; this is the part of the gap you cannot close by placement. |
Expect ranked lists to dominate the trail. They are 35.7% of content citations, the largest format by a wide margin, and in most categories the ones getting cited are published by somebody other than the brands on them.
The last row of that table deserves emphasis, because it is the part an audit cannot fix. Engines name brands without citing anything for them most of the time. When a competitor is recommended with no source behind it, there is no page to pitch and no placement to buy. Record it, size it, and be honest with whoever reads the audit that this share of the gap closes slowly or not at all.
Step 7: Turn the Trail Into a Target List
Take every third-party page that cited a competitor and did not mention you. That is your list, and it is finite, which is the whole point of doing this properly. Rank it by how often the page appeared across your runs, not by the publication's reputation, because you are optimising for what the engines actually reach for.
Then the work splits cleanly. Pages that could plausibly include you are an outreach job, and earning a place on them is a repeatable play. Pages that could never include you tell you which coverage you lack, which is a public-relations brief. Neither is a request to your content team for another blog post, and resisting that reflex is most of the value.
Comparing Yourself Fairly
Once you have appearance counts for several companies, the temptation is to turn them into a share-of-voice number. That is fine, provided you fix the three things that decide it before you compute it: the denominator, the position weighting, and how you aggregate across engines. Change any of them and the same underlying answers produce a different score.
Report per engine rather than blended. A competitor can be dominant on one engine and absent from another, and averaging that into a single figure hides the only detail that would tell you where to spend.
How Ranqo Helps
Everything above is doable by hand, and doing it once by hand is the best way to understand what the numbers mean. It stops being practical when you want it weekly, across several engines, at a sampling depth that survives the flip rate. Ranqo runs the fixed prompt set on a schedule, discovers the competitors being named alongside you rather than requiring you to list them, and records the full source trail for every answer, so step six is a report instead of an afternoon.
What it will not do is tell you why an engine recommended someone when it cited nothing. That part of the gap stays invisible to any tool, ours included, and the honest audit says so out loud. The full methodology behind the figures used here is in our 102-brand study and the paper on arXiv, including how the citation sources break down.
See who gets named instead of you
The free visibility checker runs unbranded prompts from your category against ChatGPT live and returns the competitors named alongside you, their positions, and the sources behind each answer. No signup to see the result.
Run the free checkWritten by
Nisha Kumari
Nisha Kumari is Co-Founder at Ranqo, where she leads growth strategy and client acquisition. With a background in digital marketing and financial management, she specializes in SEO, Generative Engine Optimization, and helping brands build visibility across AI platforms.
Share this article
Related articles
Your AI Competitor List Never Stops Growing
We nearly published a finding: AI names about 91 companies as the average brand's competitors, and only 18 recur. Then we sampled the same brands harder and 91 became 282, still climbing. The count was describing how often we asked, not the market. Here is the test that settles it, and what it means for every share-of-voice number.
How to Measure AI Share of Voice: The Three Decisions That Change the Number
The same prompt plays a different game on every platform -- Perplexity routinely stacks several times more citations into an answer than ChatGPT. Yet every guide hands you one formula: mentions divided by total, times 100. Share of voice is three decisions -- denominator, position weighting, aggregation -- before it's a number. This is the measurement playbook.
Your Website Is Only 2.9% of AI Citations. Here's Where the Other 97% Comes From.
Only 2.9% of the 149,912 AI citations we measured pointed to a brand's own website. Yet most AI-visibility advice still starts with 'optimize your pages.' In AI search your own domain is the floor, not the lever; the other 97% is earned on third-party pages. Here's the full source breakdown, and how to earn your way in.