A banner is circulating inside ChatGPT that invites a company to claim its business. Behind the banner sits a chart. Impressions on one line, clicks on the other, counted daily. The ChatGPT business analytics view is the most useful thing to happen to AI visibility measurement in two years, and it is going to be misread inside a week of launch.
Here is what is actually known. Juozas Kaziukenas, who runs Marketplace Pulse, posted on LinkedIn on November 6, 2026 that ChatGPT is building an analytics tool to track business mentions. Barry Schwartz wrote it up at Search Engine Roundtable the same day. The analytics sit inside a wider Business Hub, open only to companies that have verified ownership, and connecting a Shopify store is the example given. Kaziukenas says he first saw early pieces of it in August, and that the banner suggests it is close.
OpenAI has announced nothing. No help page, no name, no date, no markets. Schwartz adds one useful read of his own: this looks separate from the advertising business and aimed at organic ChatGPT answers.
So treat everything below as reporting on a feature in progress, which is exactly how you should treat it when it ships too. I am not going to pretend the detail is settled. I am going to argue about what to do with it, because the gap between a first-party number and a comparable number is where most of this year's reporting is going to go wrong.
What ChatGPT Business Analytics Will Actually Show You
ChatGPT business analytics is reported as a Business Hub view giving a verified business two daily numbers: impressions, meaning the count of times it appeared in ChatGPT responses, and clicks on links through to its website, both plotted as a chart over time.
That is it. Two series and a date axis. Read the feature list on the banner and the ambition is bigger than the chart: track how often your business is mentioned, keep your products and details current, and let a business agent answer customer questions with context you control. The analytics are one third of a product that is really about owning your own entity record inside the assistant.
The verification gate is the part worth noticing early. Ownership verification through a connected store is a clean path for a retailer and an open question for everyone else. A software company, a law firm, a manufacturer with no storefront to connect: nobody has reported how those businesses prove they are themselves. If you sell something that is not a product feed, assume the claim process is the first thing you will have to solve.
The honest version is that no SEO reporting software can tell you why a ChatGPT mention moved, because the prompt that produced it never leaves the conversation. That has been the shape of this problem since the start, and a first-party chart does not change it. It changes who owns the numerator.
Why The Same Month Can Be A 2 Percent Click Rate Or 0.85
The counting rule decides the number. One large publisher's top URL, in a single month, produced 185,000 distinct conversation impressions and 3,800 clicks on one definition, and 518,000 impressions with 4,400 clicks on another, which is the same traffic read two ways.
Those figures come from Vincent Terrasi's analysis in Search Engine Journal on December 8, 2025, citing a confidential OpenAI report, and they are the most important numbers in this story. Count one impression per conversation and the click rate is 2 percent. Count every repeat appearance inside those conversations and it falls to roughly 0.85 percent. PPC Land, which walked through the arithmetic, puts it at 0.85 against the 0.80 figure in the original write-up, and the small disagreement is itself the lesson.
| COUNTING RULE | IMPRESSIONS | CLICKS | CLICK RATE | WHAT IT IS GOOD FOR |
|---|---|---|---|---|
| One impression per distinct conversation | 185,000 | 3,800 | 2.0 percent | Estimating how many separate people saw you, which is the closer analogue to a search impression |
| Every appearance, repeats included | 518,000 | 4,400 | about 0.85 percent | Estimating total surface area inside answers, which flatters volume and depresses rate |
| Difference | 2.8x | 1.16x | 2.4x | The reason a benchmark quoted without its definition is worth nothing |
Nobody has published which rule the Business Hub chart will use. Nor has anyone reported whether a text mention, a product card, a citation link and a sidebar link all count the same, and those are four quite different experiences for a buyer. Glenn Gabe's own definition, written when he asked for a tool like this, was narrower still: impressions in AI answers based on a site's content being referenced or quoted. Appearance-based and reference-based counting are not the same metric wearing different labels.
“A number you own is not the same as a number you can compare. The first is an asset. The second needs a definition, and this one does not have a public definition yet.”
This is a rerun of an argument the industry has already had and lost once. We spent a year explaining why your mention rate dashboard has been lying to you, and the mechanism was always the same: a metric defined by whoever is selling it, quoted as though it were a standard. A first-party source removes the vendor incentive. It does not remove the definition problem, and the first vendor deck that puts a ChatGPT impression count next to a Google Search Console impression count as though they measure the same thing will be published within days of launch.
The practical guard is unglamorous. Record the definition in use on the day you first export the data, keep the export, and re-check the definition every quarter. A changed counting rule can double your impressions and halve your click rate while nothing at all happens in the business.
What ChatGPT Business Analytics Leaves Out, And Who Already Ships It
Three things are absent from the reported feature: the prompts behind the impressions, any detail about the people asking, and any competitive benchmark or ranking. Bing's webmaster tooling already ships two of those three today, for free, to anyone.
Kaziukenas is explicit about the gaps. No prompt data, no user data, no benchmark against competitors. His conclusion is the right one: the chart will not explain why mentions or clicks move, but paired with external prompt tracking it gives accurate first-party data on whether a generative engine programme is working. That pairing is the whole operating model, and it means the arrival of a first-party chart is a reason to keep your prompt tracking rather than cancel it.
| CAPABILITY | GOOGLE SEARCH CONSOLE | BING WEBMASTER TOOLS | REPORTED CHATGPT BUSINESS HUB |
|---|---|---|---|
| Impressions in AI answers | Yes, generative AI reports reached all sites on August 31, 2026 | Yes, citations reported in the AI Performance preview opened February 10, 2026 | Yes, reported as daily impressions |
| Clicks | No clicks in the generative AI reports | Not reported as a click metric | Yes, clicks through to your site |
| The queries or prompts behind it | No queries in the generative AI reports | Yes, sampled grounding queries | No, explicitly excluded |
| A competitive view | No | Yes, the October Topics Portfolio beta shows queries where rivals are cited | No, explicitly excluded |
| Breakdown by page, country, device | Yes, by page, country, device and date | Partial | Not reported |
Look at that table and the ranking of these tools by usefulness is not the ranking by attention. Bing has the only public competitive view in AI search and sampled queries alongside it, and most brands I talk to have never opened it. Google has the broadest coverage and gives you impressions with no clicks and no queries, which we unpacked in detail when Search Console's generative AI report shipped to everyone. ChatGPT is about to ship the one metric everybody actually wants, clicks, attached to the engine with the most conversational reach, and nothing to diagnose it with.
How many of the three named AI search consoles ship each capability, tallied from the reported feature sets above. Coverage count, not a quality score
Measured against what the industry asked for, this is partial progress and worth saying so plainly. Gabe published a case for an AI Search Console on July 9, 2025 with twelve features in it. PPC Land counts three of the twelve as covered by what has been reported: impressions, clicks and trends over time. Prompt data is excluded by design. API access, crawl statistics, removal tools and reporting by device and country are not mentioned at all.
How Big Is The Prize, Honestly?
ChatGPT's search-like query volume runs at roughly one fifth of Google's according to OpenAI's Adam Fry, while the largest public measurement of referral traffic puts ChatGPT at about a fifth of one percent of visits to the average site. Both numbers are true and they describe different things.
Ahrefs measured 76,000 websites in February 2026 and found ChatGPT accounted for 0.21 percent of traffic against nearly 40 percent for Google. Microsoft Clarity, looking at more than 1,200 publisher and news sites in December 2025, put AI referrals below 1 percent of visitors. Fry also says ChatGPT went from one billion to 1.2 billion users across two months. A channel can be one fifth of Google in questions asked and a rounding error in sessions delivered, because an assistant that answers well sends fewer people onward.
The conversion figure is the one to take seriously. Clarity measured AI referrals converting to sign-ups at 1.66 percent against 0.15 percent for organic search, roughly eleven times better, from a traffic base below 1 percent. Similarweb found in June 2026 that brands ChatGPT recommended were 2.5 times more likely to get a visit within seven days, and that 55.9 percent of AI-influenced visits arrived via search rather than straight from the assistant. Small, high-intent, and mostly invisible to a referral report.
Which is the real argument for claiming the property. Not that the traffic is large, but that it converts well, is growing fast, and arrives through a path your analytics attributes to somebody else. A click number from the engine is the only instrument that sees the first step of that path. The rest of the case for generative engine optimization as its own line of work has always rested on exactly this split between where the question gets asked and where the session gets recorded.
One caution on the sampling. Vendor visibility tools run prompts to produce their estimates, and the way they choose and repeat those prompts bends the result, which is the mechanism behind forced-search sampling skewing visibility data. A first-party impression count has no sampling problem. It has a definition problem instead. Different failure, same requirement: write down what you are counting.
What To Do Before The Chart Arrives
The useful work happens before launch, not after. Claim the business the day verification opens, decide now which questions the chart is allowed to answer, and keep the prompt tracking that tells you why a line moved, because the chart will not.
There is a reasonable objection to all of this, and it is worth stating properly. The feature is unannounced, the definition is unpublished, the launch could slip a quarter, and preparing for an unconfirmed dashboard is a good way to spend November on nothing. Fair. The answer is that four of the five moves above pay off whether or not the chart ever ships: verification readiness, a written definition habit, prompt tracking you already have, and a Bing account you should already own. The work is not speculative. The dashboard is.
The deeper shift is about who holds the record. For two years a brand's presence inside an assistant has been something estimated about it from outside. A claimed business profile with its own metrics makes that presence something a company holds, updates and is accountable for, which is a different relationship with the engine and a much better one. It also means the entity record inside ChatGPT becomes a thing with an owner, and owners are the ones who get asked why the number went down.
Expect the first wave of reporting to be wrong in a predictable way: a ChatGPT impression count pasted beside a Google impression count, summed into a single AI visibility figure, with no definition anywhere on the slide. Do not be the agency that ships that slide.
See where you are cited today
A free snapshot audit of your rankings and AI citations before we ever talk.
Josh leads work at the intersection of SEO and generative engines at Something Inc., helping B2B brands get ranked and cited across every major AI engine.