# ChatGPT Business Hub Metrics and Weekly Reporting

> What ChatGPT Business Hub analytics measure, why Hub totals never match GA4 or PostHog, and how to keep a private weekly log that survives the 28-day window.

Canonical: https://revnu.com/blog/chatgpt-business-hub-metrics-and-weekly-reporting
Published November 19, 2026 by George Jefferson, Co-founder and CEO, Revnu

# ChatGPT Business Hub Metrics and Weekly Reporting

ChatGPT Business Hub analytics are aggregate Insights, mentions, website clicks and discovery topics, reported over a 28-day window. They will never match your GA4, PostHog or CRM totals because they count different things in different populations. To make them useful, define each metric and its unit, keep Hub aggregates, sampled prompt tests and your own site sessions in separate columns, and write the numbers into a private log every week.

The Hub is OpenAI's "Businesses in ChatGPT" beta, and everything below about it is as checked 18 November 2026. Beta facts move, so the recheck list in each section matters as much as the definitions. The example numbers in this post come from an invented scenario, not from any real account.

## What ChatGPT Business Hub analytics measure (definition table, as checked 18 November 2026)

You sign in, claim your business, complete verification, and then open Insights. Per OpenAI's description of the beta, Insights are aggregate activity over the past 28 days, and OpenAI says businesses do not receive individual prompts or conversations (OpenAI, Businesses in ChatGPT, 2026).

| Metric | What it counts | One counterexample |
|---|---|---|
| Mentions (online business) | ChatGPT referencing your website | An answer that names your brand but never references your site does not count |
| Mentions (local business) | Your location shown as a result | A local result for your address counts even if the answer never says your brand name |
| Website clicks | Recorded clicks on links to your website in ChatGPT | A click is not a confirmed visit, a session or a conversion |
| Discovery topics | Common themes in answers that mentioned or cited your business | A topic is a theme, not a prompt and not a measure of buyer intent |

People trip on the brand-name-alone rule. For an online business, a mention is tied to the website being referenced, so a chat that says your company name from memory is invisible to this number.

Recheck before you rely on any of this: the 7D control, the data-through label, the cap on how many topics are shown, whether profile views appear, the exact click tooltip wording, any export or API, and the list of supported countries. Those are things observed in a beta interface, not guaranteed documentation.

## What each number does not tell you: units, topics and blank metrics

Each Hub number is an aggregate with an undocumented unit, so treat it as a trend signal rather than a count of people. The sources describe mentions and clicks but do not say whether a mention is one answer, one conversation or one user, and repeated references to the same page are not explained. Don't divide clicks by mentions and call the result a click-through rate for ChatGPT as a whole.

**Topics are themes, not prompts.** If "project tracking for small teams" appears as a discovery topic, you can't conclude that anyone typed those words. You can't rank topics by buying intent either. Use them to check whether the themes ChatGPT associates with you match what you sell.

**A blank is not a zero.** OpenAI says metrics may be incomplete or blank if activity has not met reporting thresholds, if data is still being collected, or if reporting is unavailable (OpenAI, Businesses in ChatGPT, 2026). No numerical threshold is published. Log a blank as "blank", never as 0, and don't plot it as a drop.

**Use limits.** OpenAI's terms limit what you may do with Insight Data: analyse and report on your own use of the service. You may not disclose it outside restricted service providers without written permission, identify people, sell it or build user profiles from it. Public posts, case studies and screenshots need OpenAI's written permission ([Businesses in ChatGPT Terms](https://openai.com/policies/businesses-in-chatgpt-terms/), section 6). Reporting to an agency acting as your service provider under equivalent restrictions is allowed. Keep the weekly log private.

## Why Hub numbers don't match your tracker: three separate data populations

Hub totals never match a tracker because they are different populations measured by different systems, and there is no user-level join between them. That answers every version of "why do my stats not match between platforms".

1. **Hub aggregates.** Real activity inside ChatGPT, counted by OpenAI, reported as aggregates with no prompts or user identities.
2. **Sampled prompt tests.** Answers you collect by deliberately running chosen prompts against AI assistants. They show how an answer reads for prompts you picked. They are synthetic samples, not real user activity, and can't be added to Hub numbers.
3. **Your own site sessions.** What GA4, PostHog or your CRM recorded after a visitor arrived, subject to consent, blockers, attribution rules and time zones.

Between Hub clicks and GA4 sessions, the usual causes of a gap are concrete:

- The visitor leaves before the analytics script runs.
- Consent settings or blockers stop tracking.
- GA4 does not start a new session for the visit.
- The two systems use different attribution rules or time zones.
- Attribution is lost or mislabelled. A visit that carries utm_source=chatgpt.com without a utm_medium can land in GA4 as chatgpt.com / (not set) under Unassigned. A visit with neither UTM nor referrer, common with app, copied-link or privacy-related traffic, can fall into Direct.

To investigate, compare GA4 session source / medium (including chatgpt.com / (not set)) with Unassigned and Direct, check landing-page sessions against Hub clicks for the same dates, and accept that missing attribution can't be rebuilt afterwards. The same logic explains most "analytics data discrepancy between tools" questions: different definitions, different capture points, different attribution.

Don't claim a capture multiplier, such as "GA4 sees 60% of Hub clicks", from a few weeks of data. And don't claim that Hub clicks caused your signups. A click is a recorded click. A signup is your system's event.

## 28-day windows, the 7D view and how to read a change week to week

Hub Insights are documented as covering the past 28 days, so each weekly reading overlaps 21 days with the last one and a single week's change is smoothed. A rise in the 28-day number can mean this week was strong, or that a weak week just fell out of the window.

If the interface shows a 7D view, treat it as an observed control to recheck, not a documented guarantee. Note the data-through label every time you copy a number, because the window may end a day or more before you look.

Here's a reading rule that holds up: change the question from "did it go up?" to "what changed in the window?". Take an invented example. Your 28-day mentions read 400 one week and 440 the next. Before celebrating, check whether the 7D view shows the new week was higher than the week that dropped out, whether the data-through date moved, and whether any metric flipped from blank to populated. Only a change that survives those checks, repeated over several weeks, earns a sentence in a report.

Mentions moving without clicks moving, or the reverse, is common and not an error. They count different events.

## Preserving history: export, API and a private weekly log (what to recheck)

The safest way to preserve history is a private weekly log you write yourself, because a rolling 28-day window overwrites itself and a beta may change what it shows. The same rule applies when you switch trackers: export before you cut over, store the raw export unchanged, and keep the definitions beside the data.

Good retention practice for any analytics source:

- **Export on a schedule, not on a crisis.** Pull data while the old tool still works.
- **Keep raw and interpreted separate.** Store the unedited export and a second sheet where you add notes.
- **Record definitions with the numbers.** A metric's unit and window are part of the data. "Mentions, 28 days, online business" is not the same as "mentions" next year.
- **Date everything.** Log the capture date and the data-through date separately.
- **Record blanks as blanks.**

For the Hub specifically, recheck whether an export or analytics API exists. As of 18 November 2026 OpenAI documents neither, so plan on manual snapshots owned by you. Copy the figures into a private sheet, a spreadsheet you control or a warehouse table you populate yourself. Joining that table to GA4, PostHog or a CRM is your own implementation, not something the Hub provides.

Keep the log private. OpenAI's terms make Insight Data confidential: share it only with service providers acting for you under equivalent restrictions, such as an agency preparing your report, and never in a public post, case study or screenshot without OpenAI's written permission.

## A weekly reporting template using an invented example

A weekly report is one row per week with three clearly separated blocks, one for each population, plus a notes column. The numbers below are invented for an imaginary project-management tool, "Plotline", and are not from any real account.

| Week of | Hub: mentions (28d) | Hub: clicks (28d) | Hub: top topic | Data-through | Prompt tests: named / run | Site: sessions from chatgpt.com | Site: signups from those sessions | Notes |
|---|---|---|---|---|---|---|---|---|
| 2 Mar | 310 | 12 | Team task tracking | 1 Mar | 3 / 10 | 9 | 1 | Clicks blank for first reading |
| 9 Mar | 340 | 15 | Team task tracking | 8 Mar | 4 / 10 | 11 | 1 | Pricing page edited 5 Mar |
| 16 Mar | 335 | 14 | Roadmap planning | 15 Mar | 4 / 10 | 10 | 2 | Topic changed |

How to use it:

1. **Fill the Hub block first**, straight from the screen, with the data-through date.
2. **Fill the prompt-test block** from the same ten prompts each week. Changing prompts breaks the series.
3. **Fill the site block** from GA4 or PostHog using one saved report. Include chatgpt.com / (not set) and Unassigned so you don't miss misattributed visits.
4. **Write one sentence per block**, then one decision. In the invented example: "Mentions flat for three weeks, topic moved to roadmap planning, so test a page aimed at roadmap planning."

Never sum the blocks. Never compute signups per mention. Note in the log that the topic shift is a theme change, not proof of demand.

## How Revnu fits: choosing and running the next approved growth experiment

Revnu is an AI agent that runs growth experiments for software businesses, and it picks up where this log stops: deciding what to do about what the numbers say. Reporting tells you that mentions are flat and a topic moved. Someone still has to choose the next move, ship it and check back.

You bring the weekly log, including whatever Hub snapshot you are permitted to use, pasted in yourself. Revnu has no native Business Hub connection, so it does not pull Hub data automatically. From that input and the work it can do for your account, it chooses one next experiment, such as writing a page for the topic that appeared, tracking how your business shows up in ChatGPT, Claude, Perplexity and Gemini answers, or running outbound and ads on the same message. It queues what needs your approval, schedules the check-back and reports the result in Slack. Once connected, Revnu also tracks AI answers and Google Search Console data, which gives you a second view next to the Hub.

For related reading, see [AI search reporting for agencies](https://revnu.com/blog/ai-search-reporting-for-agencies-what-clients-need) and [AI crawler tracking without JavaScript](https://revnu.com/blog/ai-crawler-tracking-without-javascript-what-logs-show). Pro is from $99 a month, self-serve, with no sales call.

## Start the log this week, then let Revnu pick the next experiment

The Hub gives you a trend, your prompt tests give you a sample and your tracker gives you sessions. None of them can stand in for the others, and only a log you keep yourself preserves the history. Start one row this week with the data-through date, then bring the log to Revnu at revnu.com and have it choose and run the next approved growth experiment.

## More ChatGPT Business Hub guides

- [ChatGPT Business Hub: What It Is and Who Can Use It](https://revnu.com/blog/chatgpt-business-hub-what-it-is-and-who-can-use-it)
- [How to Claim and Verify Your Business in ChatGPT](https://revnu.com/blog/how-to-claim-and-verify-your-business-in-chatgpt)
- [Turn ChatGPT Business Hub Data Into Revenue Decisions](https://revnu.com/blog/turn-chatgpt-business-hub-data-into-revenue-decisions)
- [ChatGPT Business Hub Product Feeds: CSV Setup Guide](https://revnu.com/blog/chatgpt-business-hub-product-feeds-csv-setup-guide)
- [ChatGPT Business Agents: Eligibility and Setup](https://revnu.com/blog/chatgpt-business-agents-eligibility-and-setup)

## Related Business Hub guides

- [How to Get More Website Clicks From ChatGPT](https://revnu.com/blog/how-to-get-more-website-clicks-from-chatgpt)
- [ChatGPT Discovery Topics: What They Mean and How to Act](https://revnu.com/blog/chatgpt-discovery-topics-what-they-mean-and-how-to-act)
- [How to Get More ChatGPT Mentions for Your Business](https://revnu.com/blog/how-to-get-more-chatgpt-mentions-for-your-business)
- [ChatGPT Business Hub vs Google Search Console](https://revnu.com/blog/chatgpt-business-hub-vs-google-search-console)

## Frequently asked questions

### Why do ChatGPT Business Hub numbers not match GA4?

A click may never become a session because the visitor leaves early, consent or blockers stop tracking, or attribution lands in Unassigned or Direct. Time zones and attribution rules also differ. Compare same-date landing-page sessions, not totals.

### Does a blank metric in Business Hub mean zero?

No. OpenAI says metrics may be blank or incomplete when activity has not met reporting thresholds, data is still being collected or reporting is unavailable. No numeric threshold is published. Log a blank as blank, and do not chart it as a drop to zero.

### How do I keep a history of ChatGPT Business Hub analytics?

Insights cover a rolling 28 days, so copy the numbers into a private weekly log with the capture date, the data-through date and each metric's unit. OpenAI documents no Hub export or API, so copy the figures by hand. Keep raw snapshots separate from your notes.

### What are discovery topics in Business Hub?

Discovery topics are common themes in ChatGPT answers that mentioned or cited your business. They are not individual prompts and not a direct measure of customer intent. Use them to check whether the themes ChatGPT links to you match what you sell.

### Can Revnu pull my Business Hub data automatically?

No. Revnu has no native Business Hub connection. Revnu chooses the next growth experiment, does the work within your approvals and reports back in Slack.
