Guide: AI in business › Visibility in Google and in AI model answers
An AI visibility audit is a check of whether (and how) ChatGPT, Gemini, Perplexity, Claude, Copilot and Google’s AI Mode mention your company when a customer asks about your industry, and why they mention it or leave it out. A good audit ends with a list of specific improvements and a reference point against which you will compare the result in three months’ time. We describe what it should include, how to do a basic version yourself and when outside help is useful.
In 60 seconds The AI audit in brief
- What it involves: a measurement of the brand’s presence in AI model answers on a fixed list of customer questions, together with an analysis of the sources that shape those answers.
- When it is useful: when customers ask about suppliers and compare offers before buying, especially in B2B.
- Cost: a basic audit in a spreadsheet takes a few hours of work, and tools for continuous tracking cost from about USD 99 a month (September 2026).
- Starting material: 20–40 questions phrased in your customers’ language, not in the language of your offer.
- Main pitfall: a single answer is an anecdote, so draw conclusions from repeated measurement and a group of control questions.
What is AI visibility and why does it need to be measured?
AI visibility comes down to a simple question. When a customer asks a language model for a solution to their problem, do they hear your company’s name? If so, in what company, in what tone and with accurate information? If not, whom do they hear about instead of you?
It needs to be measured because it doesn’t show up in any standard report. Google Analytics will show traffic, but not the conversation after which the customer never visited your website at all. And such conversations are on the rise: according to the Pew Research Center, users who saw an AI summary in Google clicked a regular link in 8% of visits, compared with 15% without a summary. On the Commplace® blog, 28 of the 40 phrases for which we rank in the top twenty already have an AI answer above the results (own measurement, 21 September 2026).
Visibility in ChatGPT and visibility in Google are two different things, although related. The first depends mainly on who writes about you online. The second also depends on what you publish on the website yourself. How they differ is shown in the article about that., how SEO shares the work with GEO and AEO. And if you are just starting to get to grips with AI in your company, it is better to begin with the guide “Artificial intelligence in business”, where the visibility audit is one of several elements.
What should an AI visibility audit include?
An AI audit has eight parts. Each answers a different question, and only together do they reveal the cause rather than just the symptom, so don’t skip any of them.
| Part | Question it answers |
|---|---|
| List of control questions | what customers ask before buying — 20–40 questions in their own words |
| Measurement in models | whether the brand is mentioned, in what position, with or without a link, in what tone |
| Accuracy | whether the model gives accurate information about the company: scope of services, location, prices, figures |
| Competitors | who is mentioned instead of you, and thanks to which sources |
| Sources | which pages the model cites — yours, media, directories, forums |
| Website | whether the most important pages are indexed, accessible to AI crawlers and answer questions directly |
| Fix plan | what to change on the website, where to earn mentions, what to correct in company data |
| Baseline | the result recorded so that in 60–90 days you can compare like with like |
Who is an AI audit for and when should you commission one?
An audit is useful for any company whose customers look for information before they talk to a salesperson. Most of all in these situations:
- Salespeople hear from customers that “ChatGPT recommended” a competitor. Not your company.
- The company has changed its name, scope of services or headquarters, but the models keep repeating the old data.
- You are planning to enter a new market or launch a new product and want to know what level you are starting from. How to prepare for the debut itself, we write on the topic. introducing the product to the foreign market.
- After a crisis, models summarise outdated articles. In that case, the audit ties in with crisis communication in the age of AI.
- You already pay for SEO or PR and want to check whether these activities translate into presence in AI answers.
How to check your brand’s visibility in AI yourself, step by step
How to check visibility in AI without paid tools? You can do the basic version of the audit yourself in a spreadsheet. It will take you a few hours, and you’ll get a zero point, without which any subsequent discussion about the results is guesswork. When the measurement is to be repeated every week and cover many models, the spreadsheet stops being sufficient. Then this task is taken over by Commwork®AI cockpit, which saves answers, sources, and changes in one place.
- STEP 01List the questions
Divide them into three groups: “which company does…”, “how much does … cost”, “how to solve the problem of…”. Add direct questions about the brand, e.g. “what does [name] do?” or “[name] reviews”. Also choose 5–10 control questions that you will then deliberately leave untouched.
- STEP 02Ask them in several tools
ChatGPT, Gemini, Perplexity, Copilot, Claude, a do tego Tryb AI Google. Najlepiej w oknie bez logowania albo w trybie prywatnym, żeby historia rozmów nie zmieniała wyników. Jak krok po kroku sprawdzić, czy te narzędzia polecają Twoją firmę, opisujemy w tekście jak sprawdzić, czy ChatGPT, Gemini i Google AI polecają Twoją firmę.
- STEP 03Record the results in a spreadsheet
For each question–tool pair, note: whether the brand is mentioned, in which position on the list, with or without a link, in what tone, with what errors, alongside which competitors and which sources the model refers to.
- STEP 04Open the cited sources
This is where you will find out why the model sees a competitor and not you. It could be a ranking on a portal, an industry article, a business directory or a forum thread. This is the most valuable part of the audit.
- STEP 05Check your website and robots.txt
Check whether the most important pages are indexed and whether you are blocking the OAI-SearchBot crawler. OpenAI states that sites that block it will not appear in ChatGPT search answers.
- STEP 06Record the date and tool versions
Answers can differ as early as the next day. A single result is an anecdote. Data only emerges with repeated measurement on the same questions.
How Commwork®AI transforms visibility audits into a continuous measurement
A one-time audit shows the current state, but the models’ responses change from week to week. Commwork®AI repeats the same measurement in eight AI engines, compares you with competitors, and turns every gap into a task for the owner. The Commplace® team plans the improvements, and the decision is always made by a human.
- Competitor A 31%
- Competitor B 24%
- Your brand 22%
- Competitor C 15%
- Other companies 8%
- Week 1List of 30 questions in the language of the customers
- Week 3Baseline measurement recorded with date and versions
- Week 4Source map: where models get their competitors from
- Week 8Incorrect company data corrected at the source
- Week 12Control measurement and comparison with the database
- 01List of 30 customer questions
- 02Measurement in 8 AI engines
- 03Map of sources and gaps
- 04Fix plan
- 05Measurement after 90 days
- It asks the same list of customer questions in eight AI engines and records who the models are swapping with, in what tone and based on what sources.
- It shows what’s missing about you and what’s missing about the competition, and turns every gap into a task with a deadline and a KPI.
- It distinguishes the effect from the case: control questions that you don't touch on act like a comparison group. Why the media sources matter so much here, we explain in the text about AI PR and visibility in ChatGPT.
The view reproduced is modeled after the dashboards provided to customers. The names, numbers, and terms are examples and show the way the system works, not the results of a specific company.
How to read results from ChatGPT and other models
Models differ in how they cite, so don’t measure them all with the same yardstick. According to the Muck Rack study “What Is AI Reading?” from May 2026, ChatGPT gives sources in 96% of answers, but only five on average. Gemini cites in 82% of answers, with eight sources on average. Claude is the most selective: it cites in 55% of answers, but when it does, it gives 13 sources on average.
What does this mean for the audit? In ChatGPT, you are competing for one of a few places, so presence in the strongest sources is what counts. In Claude and Gemini, the list is longer, so it is easier to get onto it but harder to stand out. The same study shows that 84% of citations are earned media, and paid content accounts for 0.3%. That is why press articles so often end up in the “sources” column, and why AI visibility is also a topic for the PR department, which we expand on in the article PR AI.
Google is a separate measure. Search Console now has a separate report for generative features, with impressions of your pages in AI Overviews and in Google’s AI Mode (without clicks), and since the end of August 2026 it has been available globally. You can recognise traffic from ChatGPT in your analytics by the tag utm_source=chatgpt.com.
How much does an AI visibility audit cost and what determines the budget?
The budget depends on how many questions and models the measurement covers, how often you repeat it and whether the audit ends with a diagnosis alone or also with an improvement plan including source analysis. Below are public tool prices (as of September 2026) and the factors that change the cost.
Before you spend money on a tool, calculate how much the lack of knowledge costs you. If one client from the quote request is worth more to the company than a one-year subscription for $99 per month, ongoing visibility monitoring in AI pays off with the first conversation recovered. If not, stick with the spreadsheet and repeat it once a quarter.
Set the decision threshold before the first measurement. For example: if the brand appears in less than one in three purchase questions or the model provides incorrect information about the company, the audit moves into a plan of corrective actions. If you want someone to do this for you and immediately point out the sources of improvement, see how it works auditing and positioning in AI at Commplace.
| Item | Cost or factor | When it is enough |
|---|---|---|
| Manual audit in a spreadsheet | PLN 0 for tools, a few hours of work | first measurement, 20–40 questions, once a quarter |
| Semrush AI Visibility Toolkit | from USD 99 a month per domain, 25 custom questions, billed annually | continuous tracking of ChatGPT, Google AI, Gemini and Perplexity |
| Ahrefs Brand Radar AI | from USD 199 a month as an add-on to an Ahrefs subscription | large-scale analysis of brand and competitor mentions in AI answers |
| Number of questions and models | more questions = more work or a more expensive plan | 20 questions to start with, 40+ for multiple product lines |
| Source analysis and improvement plan | analyst time, usually the largest part of the work | when the result is meant to lead to a decision, not just a report |
How to increase your brand’s visibility in AI after the audit
How to increase brand visibility in ChatGPT and other models? The audit will reveal gaps. The plan for improvements usually falls into three directions. The first is the page: the answer in the first sentence of each important subpage, the numbers from the source and the date, the author’s signature, the order in the indexing. The second is external sources, that is, expert publications in industry media (each Press mention This is another independent source for the model), presence in listings and consistent company profiles in catalogs. The third is data: everywhere the company is described, the same company name, the same address, and the same range of services.
Don’t try to boost your brand’s visibility in ChatGPT with tricks such as mass-produced texts, hidden instructions for models or bought posts posing as independent reviews. Models are getting better at spotting this. More importantly, though, you are risking your reputation among people. The honest route is slower, but its effects last. We describe what works in ChatGPT itself in the article on how ChatGPT chooses whom to recommend, and we cover minor technical elements when discussing the llms.txt file.
Visibility audit in AI in 90 days using a typical company example
The scenario below is illustrative. It does not describe any particular client; it simply shows how an audit turns into work and what rhythm measurement has.
Week 1–3. Let's say a B2B service company compiles 30 questions that customers ask before making a purchase, and adds eight control questions. The measurement takes place in several models, and the results are fed into a spreadsheet with a date stamp. The team quickly sees how to check the visibility in ChatGPT in a repeatable way: the same questions, the same private mode, the same day of the week.
Week 4–8. The column with sources provides the most work. It turns out that the models are based on two industry sets and one old company catalog. The team improves the data in the catalog, prepares an expert commentary for the media with their own numbers, and formats three subpages so that the answer is in the first sentence. No one bothers to answer any follow-up questions.
Week 9–12. The second measure of visibility in AI on the same list. The change only matters when the group of questions covered by the improvements grows more clearly than the control group. If both grow equally, it’s more likely a change in the model itself than an effect of your work.
Checklist before you start:
- The questions are written in the customer's language, not in the names from the offer.
- You have a separate set of control questions that you don't answer.
- Each result has a date, a tool, and information whether logging was performed.
- For each answer you open the cited sources, and you don’t just read the text.
- You know who in the company is updating the data in the catalogs and profiles.
- You know the date of the control measurement and it is in your calendar.
What an AI audit won’t tell you
It is good to know the limits of measurement so as not to draw conclusions the data cannot support:
- how many people saw an answer featuring your brand (apart from Google’s report, there is no public counter of impressions in models),
- whether an increase after the changes is due to your actions; for that you need a group of control questions that you leave untouched (correlation is not causation),
- what the answer looks like for a specific customer, because models adapt to the user’s chat history and location,
- how many sales presence in the answers generated; this has to be researched separately, e.g. with the question “how did you hear about us?” in the contact form.
Which agencies work on AI visibility and how to choose
AI visibility is currently handled by SEO agencies, PR agencies and specialised GEO teams. For a Polish company, the most important thing is that the agency combines work on the website with content writing and media relations, because an audit almost always reveals gaps in all three areas. Ask for a sample baseline measurement report and ask how the agency will distinguish its effect from chance. We have listed the remaining questions for a first meeting in the article on how to choose a GEO or AI agency.
If you are looking for a way to check a brand's visibility in AI using the agency's hands rather than your own, ask these questions at the first meeting:
We describe how we measure brands’ presence in AI answers on the page the Commplace method. The classic communication auditFirst the facts, then the decisions. Visibility audit across eight AI engines, including Google AI Mode, with a plan for improvements, makes us GEO agency. You can book a first, free conversation via the contact form.
- What control questions will you leave unchanged and how will you distinguish the effect from the random variation?
- In how many AI engines do you measure and do you record the date and mode of each measurement?
- Does the report show the sources on which the model for the answer to the question about competition is based?
- Who among you writes the content, and who talks to industry media?
- What will I get after 90 days if the result does not change?
Where a first audit most often goes astray
Model answers fluctuate from day to day. If you base a decision on a single screenshot, you risk spending the budget on fixing a problem that doesn’t exist.
Customers don’t ask about “comprehensive intralogistics solutions”. They ask: “who can build racking for a warehouse?”. An audit on the wrong questions will give a falsely good or falsely bad result.
A simple “brand present / brand absent” list doesn’t tell you what to do. If you don’t open the cited pages, you won’t find out where your competitor’s advantage comes from.
Without them, every increase looks like a success. Yet models change on their own, and you may attribute to yourself the effect of an update you had no influence on.
If the model cites only external rankings and articles, rewriting your own website won’t help much. It costs a quarter of work with no effect.
Questions companies ask us before their first measurement
How do you check your brand’s visibility in ChatGPT?
Prepare 20–40 questions that customers ask and enter them in ChatGPT in a logged-out window. Note whether the brand is mentioned, with what source and in what tone, and repeat the measurement a month later. You can check traffic from ChatGPT to your website in analytics by the tag utm_source=chatgpt.com.
How much does an AI visibility audit cost?
A basic do-it-yourself audit takes a few hours of work and costs nothing for tools. Tools for continuous tracking cost from around USD 99 a month (Semrush) or USD 199 a month (the Ahrefs Brand Radar AI add-on), as of September 2026. The price of an audit with an agency depends on the number of questions and models and on the scope of the source analysis. We carry out an audit with analysis of sources, competitors and an improvement plan as part of a diagnosis, which with us costs from PLN 8,000 net; the first conversation is free of charge (as of 24 September 2026).
How often should you repeat an AI visibility audit?
A full audit once a quarter, a shortened measurement on the same questions once a month. More often makes no sense, because model answers fluctuate from day to day and it is easy to mistake a random change for a trend.
How does an AI audit differ from an SEO audit?
An SEO audit checks the website and its search engine rankings. An AI visibility audit checks what models say about the company and looks for causes beyond the website too — in the media and directories, on review sites and in the consistency of company data. One complements the other.
How do you increase your brand’s visibility in AI?
Improve your pages so that they answer customers’ questions directly, earn coverage in independent industry media and make your company data consistent across the web. According to Muck Rack, 84% of citations in AI answers come from earned media. Check the effect with a repeat measurement that includes a group of control questions.
Does visibility in Google mean visibility in AI?
Partly. AI answers in Google use the same index as regular results, so good SEO helps. ChatGPT, Claude and Perplexity, however, also draw on other sources, which is why a company can rank high in Google and still not appear in their answers.
Which agencies work on AI visibility?
SEO agencies, PR agencies and specialised GEO teams. When choosing, check whether the agency can both improve the website and earn media coverage, and ask for a sample baseline measurement report. Avoid offers that guarantee a specific place in model answers.
Sources
- SOURCE
- SOURCESearch Console Help: performance report for generative features · as of 24 September 2026
- SOURCESearch Engine Land: AI reports in Search Console available globally · 31 August 2026
- SOURCE
- SOURCEMuck Rack: “What Is AI Reading?” — May 2026 edition · 7 May 2026
- SOURCE
- SOURCEOpenAI: Publishers and Developers FAQ (OAI-SearchBot, utm_source=chatgpt.com) · as of 24 September 2026
- SOURCESemrush: AI Visibility Toolkit pricing · as of 24 September 2026
- SOURCEAhrefs: pricing and the Brand Radar AI add-on · as of 24 September 2026
Read next
We will perform a visibility audit across eight AI engines, including Google AI Mode: your customers’ questions, source and competitor analysis, a plan for improvements, and a baseline for measurement in 90 days.
Sebastian Kopiej, President of Commplace®. In public relations since 1996. Text prepared with the help of AI tools and fact-checked by the author. Data current as of September 29, 2026.