Will ChatGPT Recommend Your Company When a Customer Asks About Your Category? The 20-Query Test

I see Google traffic declining, CPC rising, and the board asking why we're paying for content. I don't know whether our company even shows up when a customer asks ChatGPT about our category, and I'm afraid we've fallen off the shortlist before anyone ever contacted us. I heard this sentence this quarter from more than a dozen executives, and from myself, staring at the organic sessions chart in July.

The old question was: which keywords do we rank for? The new question is: are we even in the answer when a buyer asks about the category? This shift sounds innocent, but it dismantles the entire existing reporting model, because the answer isn't in Search Console — it's in a chat window that nobody measures.

Zero-click search and the loss of control over the pipeline

SparkToro's 2025 data shows that 64 percent of Google searches end without any click, because the user gets the answer directly in an AI Overview or an expanded snippet. Bain's research shows that after Google introduced AI-generated summaries, click-through rate in B2B software categories dropped by as much as 30 percent. Cost per click in the B2B segment rose 24-29 percent year over year, and 51 percent of B2B buyers start their research with AI assistants before ever reaching a brand's website.

This combination of numbers explains why declining B2B organic traffic looks like a catastrophe, even though it's really a change in what a search engine does. Google is no longer a mechanism that drives traffic to pages; it's becoming a mechanism that delivers instant answers within the result itself. If a team reports effectiveness through organic sessions and form-based MQLs, this shift looks like an SEO collapse, and the board draws the rational conclusion: content isn't working, cut the budget.

In this environment, the traditional model of "content, rankings, clicks, leads" no longer describes the real impact of content on sales. CMOs and VPs of Sales lose a clear reading of whether the brand even exists at the moment a buyer forms their first question about the category. Content-to-sales attribution splits into two parts: the part visible in UTMs, and the part that happens inside a model's answer before anyone ever visits the site.

The fair counter-case: Commercetools and stable conversions

Before anyone panics and declares SEO dead, it's worth looking at a case described by Digiday. Commercetools' CMO recorded a 20 percent drop in CTR from Google after AI Overviews rolled out, while conversions stayed at the same level. This result suggests that zero-click mostly eats into the top of the funnel — informational traffic from people who wouldn't have bought in that session anyway.

A drop in CTR says nothing about how many buyers saw the brand in ChatGPT, and that's the gap this test closes. The operational takeaway for CMOs is this: the quality of remaining clicks rises, because they increasingly come from people with a partially formed preference who are looking for a specific brand. The problem arises when the brand isn't in the AI answer at all, because then there's nothing to convert.

Brand visibility in AI answers as a measurable indicator

I define brand visibility in AI answers as the frequency and manner in which a brand is mentioned or cited in models' answers to a defined set of buyer queries, at a given moment and in a given market. Classic SEO answers the question of ranking; AI visibility answers the question of citation — determined by the models' training knowledge, the current web index, and trust in sources.

The absence of a mention in an answer to a query like "best B2B lead generation automation platforms" means that, from the buyer's perspective, the brand effectively ceases to exist in the category. This is the mechanism behind a company falling off the shortlist: you don't lose the lead in the CRM — you lose it the moment the model builds its list of recommendations and you're not on it.

How to check whether ChatGPT recommends my company

How do you check whether ChatGPT recommends your company? Run 20 category questions three times on the same day, count in how many of the 60 answers the brand appears, and log the result with a date — that's your AI visibility baseline. This baseline is the entry point for a conversation with the board, because it turns an anecdote into a number that can be compared week over week and month over month.

How ChatGPT, Perplexity, and Gemini build their answers

ChatGPT, Perplexity, and Gemini don't work like a search engine, because each has a different mechanism for trusting sources, and that difference determines where you need to work on visibility.

ChatGPT builds answers based on training knowledge, reference content, industry articles, and data from databases such as Crunchbase or LinkedIn. Wikipedia, an About page, and entity consistency across company profiles are fuel that lets the model build a picture of an organization and its position in the category. If these sources say different things or stay silent, the model has nothing to build a recommendation from.

Perplexity shows the user a list of sources used to generate the answer, so brand visibility here is tied to whether a page is cited as a source. In Perplexity you measure citations, not mentions, because the mere presence of a name without a link to a source doesn't translate into a real signal.

Gemini relies heavily on the Google Knowledge Graph and signals from YouTube and classic SEO. Visibility in Gemini is more closely tied to domain authority and presence in structured data, so the Schema Graph and entity order do more of the work here than any single post.

A professional audit runs the same basket of questions across three or four engines, and also classifies sentiment as "not recommended," because a negative characterization can hurt more than no mention at all. llms.txt is not a visibility lever — models draw on trusted sources, not a file that nobody queries.

The "20-Query Test" protocol — step by step

In my own test, I measured 12 out of 20 category queries, as of March 14, 2026. The biggest surprise: the brand didn't appear even once in category questions — it only appeared in branded queries. That sentence cost me three hours of work and changed the way I talk about content budget.

Step 1: selecting the 20 category questions

The first step is selecting twenty representative questions that mirror real purchase intent in the segment. The questions should cover both educational queries and those closer to a purchase decision, because only then can you see whether the brand exists at the research stage and at the shortlist stage.

The practical list I run looks like this: five category questions ("what is B2B lead generation automation"), five comparison questions ("tool X vs. tool Y"), five recommendation questions ("best B2B lead automation platforms"), and five problem questions ("how to reduce cost per lead in B2B"). Each of these is a buyer category question that genuinely comes up in a conversation with an AI assistant.

Step 2: running it three times and collecting answers

I run each question three times on the same day to get 60 answers and see the variance. I copy the answers into a spreadsheet along with the date, time, and model name, because a baseline without a date doesn't exist and can't be compared to anything.

Step 3: metrics you log in the spreadsheet

• Citation rate — I calculated it as the share of answers in which the brand appears at least once. With 60 answers and 12 mentions, the citation rate is 20 percent.
• AI share of voice — the number of brand citations divided by the total number of samples, multiplied by 100. This metric shows the brand's share across the entire basket of questions.
• Competitor citation gap — for every prompt where the brand didn't appear, I note who appeared instead. This produces a table showing whether the problem is total invisibility or low prominence behind better-known players.
• Position-in-answer — I check whether the brand appears at the beginning, middle, or end of the answer, because position in the text affects whether the buyer even reaches it.
• Sentiment — I mark it as positive, neutral, or negative, and I check reproducibility by comparing variance between runs of the same prompt.
• Reproducibility — I measure how many times the same prompt produces different answers, because a model that recommends the brand once and not another time is a less reliable witness than a stable model.

Step 4: the competitor citation gap table

For every prompt where the brand didn't appear, I note who appeared instead — this produces the competitor citation gap table. In practice, this table is the most-read part of the report, because the board wants to know who we're missing next to, not just how often we show up.

Step 5: logging the result with a date and status marker

I log the result with a date and status marker, for example 12 out of 20 queries, as of March 14, 2026. This format is part of the discipline of measurement, because it shows that AI visibility is a recurring indicator, not a one-off audit.

The visibility baseline as a foundation for board decisions

The results of the protocol give the CMO hard evidence for a board report. The board gets a repeatable reading: in how many queries the brand appears as a recommended provider, alongside which competitors, and with what sentiment. This baseline lets you shift the focus from traffic volume to the quality of brand visibility in AI answers.

From a traffic-based to a trust-based model

PresenceKit frames this shift as a move from a "traffic-based" model to a "trust-based" model. In the old model, the cycle looked like this: content, rankings, traffic, leads, pipeline. In the new model: content, authority, awareness, preference, pipeline. Zero-click means that much of content's impact is dark — unattributed to any click — but it still shapes which brands make it onto the shortlist.

Five zero-click metrics to put in your report

StriveLabs proposes a set of five metrics that connect content's impact to the pipeline even when most searches end in an AI answer with no click.

• Branded search velocity — the acceleration of branded searches, i.e. whether more people are typing the company name into Google after encountering the content.
• AI citation share — the brand's share of AI citations, i.e. how often we're the source or recommendation in model answers.
• SERP feature share — the share of SERP elements, such as AI Overview and expanded snippets, that capture attention before a click.
• AI-referrer direct traffic — direct traffic from AI assistants, i.e. visits from links inserted into model answers.
• Page presence in closed sales opportunities — the number of pages that appear in the paths of closed deals, showing the real impact of content on the pipeline.

The baseline from the 20-query test is the entry point into this layer, because without it you don't know whether you're even in the game, and with it you can track how citations translate into branded searches and direct traffic.

The numbers that change the conversation with the board

Brands cited in AI Overview achieve a 35 percent higher organic CTR and a 91 percent higher CTR in paid channels than uncited competitors on the same queries. AI visibility thus becomes a recurring metric with a date in the spreadsheet — checkable weekly or monthly.

At AGAPE, I check this baseline every week, because until you have a dated baseline, every conversation about content budget is guesswork. The 20-Query Test takes three hours and delivers something you can't buy for any amount of money: a reading of your presence in AI answers that you can show to the board and defend.