AI Search Visibility Checker: Measure Where You Show Up

An AI search visibility checker is any method that records whether, where and how often your brand appears inside AI-generated answers for a defined set of buyer queries. It can be a paid dashboard or a spreadsheet and a browser. What matters is that the check is repeatable, dated and evidenced, because an answer you cannot reproduce is an anecdote.

Most B2B teams start this by typing their company name into ChatGPT, seeing it mentioned, and calling it done. That test tells you almost nothing. Your buyers do not ask about you. They ask what a rotary lobe pump costs, which hoist supplier ships to Rotterdam, or how to size a screening plant. The checker exists to measure those questions instead.

What a visibility checker actually measures

Strip away the dashboards and four quantities matter. Citation presence is whether your domain is named or linked in the answer. Citation share is how often you appear across your query panel versus competitors. Position is how early your brand is mentioned relative to others. Sentiment and accuracy is whether the model describes you correctly, which is a different problem from being absent.

A fifth measure is quieter but often the most useful: which page of yours got cited. If the model keeps pulling a 2019 blog post instead of your product page, your visibility is real but pointed at the wrong asset. You can fix that with content work. You cannot fix it by refreshing a dashboard.

Tools automate the collection of these numbers. They cannot decide what you should be measuring. That decision is yours, and it is the part that determines whether the report changes anything.

How to run a manual check in ChatGPT search

The manual check is the reference standard. If a tool disagrees with it, trust the manual check and investigate the tool. Here is the procedure our team uses, and it takes about forty minutes for a panel of twenty queries.

1. Use a clean, logged-out session. Open ChatGPT in a private browser window and do not sign in. A logged-in account carries memory and personalization that a new buyer does not have. According to OpenAI's published help documentation, logged-out sessions behave differently from personalized ones, which is exactly why the distinction matters for measurement.
2. Turn on search mode. The answer must come from live web retrieval, not from parameters stored in the model. If the response arrives without any web sources, you are measuring training data, not visibility.
3. Ask one query per conversation. Starting a fresh chat for each query prevents earlier questions from shaping later answers. It is slower. It is also the only way to compare results week over week.
4. Screenshot the full answer. Capture the query, the date, the response text and the cited sources in one image. Name the file with the date and query ID so it can be retrieved months later.
5. Log four fields. Cited yes or no, your position in the mention order, the source URL the model used, and whether the description of your product is accurate.
6. Repeat on the same weekday each month. Same panel, same browser conditions, same logging format. Consistency is what turns screenshots into a trend line.

One caution. ChatGPT answers either from live web search, which optimization can influence, or from knowledge stored in the model without web access, which currently cannot be optimized at all. If your query returns no citations, you are in the second category and no amount of content work will move that specific answer today.

[free] Not sure whether ChatGPT cites you for these queries today? We check and reply within 24 hours. Get a Free AI Visibility Audit

Building a query panel that reflects real buying

A query panel is a fixed list of questions you will track every month. Twenty to forty queries is a workable range for a mid-sized exporter. Fewer than fifteen and a single result swings your percentage wildly. More than sixty and you will stop running the check by month three.

Build it in four bands. First, category questions (what is a hydraulic winch used for). Second, comparison questions (electric hoist versus chain hoist for port work). Third, supplier-discovery questions (who manufactures explosion-proof cranes in Europe). Fourth, brand questions (is your company name reliable, who owns it). The first three bands are where new buyers live.

Pull the actual wording from your sales inbox, your live chat logs and your Google Search Console queries. Buyers phrase things differently from marketers. A query panel built from internal assumptions will flatter you and teach you nothing.

Weight the panel if your market is concentrated. A manufacturer selling mainly into Germany and Canada should carry more German and Canadian phrasing than generic English questions, because the model's answer changes with the market implied by the query.

Manual checks versus tools: what each is good for

Tools earn their place once you have more than one brand, more than one market or more than one language to track. Below is how the two approaches compare in practice, based on how our team runs client programs.

Our working rule is simple. Run manual checks until the panel is stable and you trust the numbers, then automate the collection and keep one manual spot check per month as a control. Automation that you cannot verify is just a nicer-looking guess.

If you want the diagnostic version of this exercise, an AI visibility audit maps your current citations against competitor citations before you commit to a tracking routine.

Setting a baseline and reading the trend

Your first run is a baseline, not a score. Expect it to be uncomfortable. In one RAGSEO client program (client anonymized), a lifting equipment manufacturer appeared in less than 1% of AI-generated results before the project began. After the work, AI-engine-driven inquiries reached 186, which was 35% of all inquiries, with 62% of those coming from Europe and North America at a 28% higher conversion rate than traditional channels, and the brand held a top 3 position in AI answers for core queries. That is the shape of a mature program, not a first month.

Read the trend in three layers. Citation rate across the panel tells you whether you are being found at all. Source URL tells you which pages the models prefer. Accuracy tells you whether being found is helping or hurting. A brand cited in the wrong context can lose deals quietly.

Give the trend at least two full quarters before drawing conclusions. AI answers shift with index updates, model releases and competitor publishing. A single bad month is noise. Three consecutive flat months with rising competitor citations is a signal.

If you are still deciding how this fits your overall plan, the mechanics of retrieval and citation are worth understanding first: how AI search works explains why some pages get quoted and others do not.

Where visibility work goes wrong

The most common failure is measuring brand queries and calling it visibility. Nobody discovers a supplier by asking about the supplier. Track discovery queries, and treat brand queries as a separate accuracy check.

The second failure is optimizing for a tool's score instead of the answer. If a dashboard says you improved while your sales team still hears "we found you on Google, not on ChatGPT," the dashboard is measuring the wrong thing.

The third is expecting ChatGPT work to move every engine identically. Optimizing for ChatGPT tends to help visibility in Gemini and Grok too, because they reference public web content, but each model has its own mechanism. We evaluate against ChatGPT search results specifically, and we say so rather than promising a uniform lift everywhere.

Finally, do not treat publication as the finish line. Content you publish may also enter future models' training data over time, which is a slow compounding effect, not a switch you flip. Programs that keep publishing past month six are the ones that hold their citations when a model updates.

Once the baseline is honest, the work is mostly editorial: build pages that answer the query directly, mark them up so machines can parse them, and distribute corroborating content so more than one source says the same thing. That workflow is what our AI search optimization services cover, and the monthly cadence is set out in the 12-month visibility plan.

Frequently asked questions

Can I check AI visibility without paying for a tool?

Yes, and for a single brand it is often better. A logged-out ChatGPT session, a fixed panel of twenty to forty buyer queries, and a spreadsheet with four fields will give you a defensible baseline in an afternoon. You lose scale and history, not accuracy.

Why does my brand show up when I am logged in but not when I am logged out?

Logged-in sessions carry memory and personalization, so the model may reference your earlier conversations or known interests. OpenAI's help documentation describes this difference. Buyers arrive without that context, so the logged-out result is the one that reflects commercial reality.

How often should I re-run the check?

Monthly is enough for most B2B manufacturers and exporters. Weekly checks add noise without adding decisions, because answers shift with index updates rather than on a predictable schedule. Re-run immediately after a major model release if you want to see the short-term effect.

Does a good AI visibility score mean more inquiries?

Not automatically. Citation is a precondition for being considered, not a guarantee of conversion. The programs that convert pair strong citation rates with pages that answer commercial questions clearly, including pricing logic, specifications and shipping terms.