Answer
How do you audit whether AI assistants can find your business?
Run the buying-intent questions your customers actually ask through ChatGPT, Gemini, Perplexity and Google AI Overviews, and record who gets named. Then check the same queries in ordinary search, confirm which of your pages a crawler can retrieve, and compare the wording across your site, profile and listings.
The short version
- Build a fixed list of buying-intent questions in your buyers' own wording, not your service names.
- Run them across several assistants, logged out, with the right location set, and record which businesses are named.
- Run the same questions in ordinary search and note where the two disagree.
- Verify a crawler can actually fetch and read your key pages.
- Sweep your site, Google Business Profile and listings for wording that disagrees with itself.
- Keep the question set fixed so the next run is comparable. A baseline you cannot repeat is not a measurement.
Step one: write the question list
This is the step that decides whether the audit is worth anything, and the one most people get wrong by using their own vocabulary.
You need the questions a buyer asks in the last few minutes before choosing. Not "aircon services Jakarta" but "who can fix an aircon today in South Jakarta", "is it cheaper to repair or replace an old aircon", "which aircon service is open on Sunday". Include comparisons, because comparison questions are where assistants most often name a specific set of businesses. Include price questions, which are the ones buyers ask and businesses most avoid answering.
Ten to twenty questions is enough. Write them down and do not change them, because the point is to re-run exactly these later.
Step two: run the prompt panel
Ask each question in each assistant. ChatGPT, Gemini, Perplexity and Google AI Overviews is a reasonable panel; add whatever your market uses.
Three rules make the results mean something. Log out, or use a private window, so your own history does not bias the answer toward you. Set the location you actually sell into, not the one you happen to be sitting in. Ask each question in each engine more than once, because retrieval is not deterministic and a single run is anecdote.
For each question, record: were you named, were you named accurately, and who else was named. That third column is the valuable one and the one people forget to write down. The named competitors are the actual competitive set as the engines see it, which is frequently not the set the owner would have listed.
Step three: check retrieval
An invisible page explains most bad results, so verify it rather than assuming.
Confirm your key pages return quickly to a plain request, and that the content you care about is present without JavaScript running. Check that prices, service lists and hours are real text and not baked into an image. Confirm robots rules are not blocking more than intended. Check that each service has its own URL, because a single-page site gives retrieval nothing specific to return.
Step four: the consistency sweep
Put your website, your Google Business Profile and every third-party listing side by side and compare them field by field: business name, address format, hours, service areas, categories, and the exact wording of every service.
Every disagreement you find is a reason an assistant cannot state a fact about you confidently. This is the least interesting step and it is usually where the problem is.
What a finished audit looks like
| Output | What it tells you |
|---|---|
| Question list with named businesses | Whether you are in the conversation at all |
| Accuracy notes per mention | Whether being named is helping or hurting |
| Competitor set, by name | Who is currently capturing your demand |
| Search vs assistant comparison | Whether this is a ranking problem or a citation problem |
| Retrieval check | Whether anything else on the list can even work |
| Consistency table | The cheapest fixes, usually the largest |
| A fixed baseline | The thing every later run is measured against |
What it costs
| What | Price | Timeline |
|---|---|---|
| Visibility audit | Rp 710.000, free with a website build | ~3 days |
| Competitor analysis | Rp 1.250.000, covers 3-5 competitors | ~1 week |
| Google Business Profile optimisation | Rp 1.520.000 per month, minimum 3 months | 1-2 weeks to set up |
| Workflow automations | Quoted per project, add-on only | 1-2 weeks |
| Website, new or revamped | from Rp 7.550.000 | 2-3 weeks |
Two bundles exist, priced below the sum of their parts: website plus audit at Rp 7.550.000, which is the audit included free with a build, and everything except automation at Rp 8.540.000 one-time plus the Rp 1.520.000 monthly profile fee. Only the one-time work is discounted in a bundle; the monthly fee is the same either way.
The website figure covers a revamp or company profile up to about five pages, and a custom multi-page build is quoted per project. The profile work is monthly because a Google Business Profile does not stay fixed: hours change, categories get reshuffled, and reviews and questions keep arriving. A one-off cleanup decays within months, so it is not sold that way.
The audit is where everything starts: whether search engines and AI assistants name your business for buying-intent questions, whether they describe it correctly, and which businesses get named instead of you, by name. Plus a retrieval and consistency test across your site, your Google Business Profile and your other listings. About three days.
When this is a bad idea to do
Do not audit once and treat it as a result. A single run across a handful of questions is noise. Without a fixed question set and a repeat, you cannot tell improvement from randomness, and you will attribute both to whatever you changed.
Do not audit while logged into accounts the assistants have memory of. Personalisation will show you a flattering picture of your own visibility, which is worse than no information.
Do not audit before you can act on it. If you have no capacity to fix a retrieval problem or reconcile your listings in the next few months, the findings go stale and you have bought a document.
Do not expand the panel to every engine and phrasing you can think of. A huge unrepeatable sweep is less useful than twenty questions you will genuinely re-run next month.
Common questions
- Can I do this myself?
- Yes, and you should at least try it once. Write down ten questions a buyer asks before choosing, ask each assistant, and record who is named. The discipline is in using your buyers' wording rather than your own, and in keeping the questions fixed so later runs compare.
- Why do I get different answers each time I ask?
- Because retrieval varies by phrasing, location, session history and the model's own randomness. That is why a single spot check proves nothing. You need a fixed question set run across engines and repeated, so a change in the pattern is distinguishable from noise.
- Should I log out or use a private window?
- Yes. Personalisation and memory of your previous chats will skew results toward your own business, which is the one bias guaranteed to make an audit useless. Log out, clear context, and set the location you actually sell into.
- How often should it be repeated?
- Monthly is enough for most businesses, and after any significant change to your site, profile or listings. The value is in the comparison against a baseline, so the first run matters mostly as the thing later runs are measured against.