OpenAI now publishes an official profile and an engagement dashboard for every business that claims one, which removes most of the reason to pay a third-party ChatGPT rank tracker. The authoritative signal is first-party, free during beta, and tied to real user activity rather than synthetic prompt sampling.
Third-party tools that pitch "AI visibility" leaderboards charge between $2,000 and $15,000 a month to approximate what the Business Hub reports directly, on a surface where the same prompt returns a different ranked list on nearly every run. The arithmetic stops working once the first-party option is live and the research team understands what it actually measures.
Claim the Business Profile Before Buying Anything Else
OpenAI opened the Businesses in ChatGPT beta on November 10, 2026, with a Business Hub where companies claim profiles, update information, and see aggregated engagement data for a service that reaches 1.2 billion people each week. The claim flow is documented in OpenAI's help center and takes a verified domain, an admin email, and an editable profile.
A claimed profile controls what ChatGPT returns when it surfaces the business in an answer:
- Name, logo, short description, and canonical URL
- Category, hours, and contact details exposed to Business Agents acting on behalf of users
- Linked properties, so cross-referenced entities roll up to the same record
- Admin seats and read access to the Insights dashboard
Nothing on the profile is paid. Sponsored Agents and ChatGPT ads were announced alongside the Hub but are not live during the beta, so the claim window is genuinely free and the only immediate cost is the engineering time to verify the domain.
What the Business Hub Insights Dashboard Actually Reports
The Insights tab shows aggregated mentions and website clicks over 7-day or 28-day windows. Clicks counted are clicks on links inside ChatGPT, not completed visits. Individual user conversations and Business Agent chats are never exposed to the business, which matters for both privacy and expectations: the Hub is a volume meter, not a session log.
That meter is attached to a serious channel. OpenAI's own economic research puts shopping prompts at about 50 million per day out of roughly 2.5 billion daily prompts, and referral traffic to specific retailers now reaches shares that would have been impossible a year earlier.

Why Third-Party Rank Trackers Return Different Numbers Every Run
ChatGPT answers are non-deterministic by design. A tracker that queries a prompt once and records the ordered brand list is reporting a sample of size one, drawn from a distribution the tracker cannot see. A SparkToro and Gumshoe study of 2,961 prompts across ChatGPT, Claude, and Google's AI features found the same brand list came back less than 1% of the time and the same list in the same order less than 0.1% of the time. The researchers' summary is blunt: any tool that publishes a ranking position in AI search is selling a number that does not exist as a stable quantity.
Volatility is only one failure mode. Trackers also miss tool calls (the retrieval and product-lookup steps the model runs behind the scenes), lack the personalization context a real user carries into the session, and cannot see when a Business Agent, rather than the base model, is the component returning the answer. The reproducibility problem is independent of prompt wording.
Rank tracking breaks down in AI search for structural reasons, not product-maturity reasons, and the Business Hub does not fix that for third-party vendors because they still cannot see the underlying session.
The Per-Prompt Pricing Model Does Not Scale
External trackers bill by prompt, with no cache and no batching discount. A plan that tracks 300 branded prompts across four models daily runs about 36,000 inference calls per month before any statistical replication. At roughly $0.04 to $0.08 per prompt once the vendor's margin, routing, and scraping surcharges are added, the floor for a modest brand program sits around $1,500 to $3,000 per month and climbs linearly with every prompt added. Competitor and category coverage pushes most enterprise plans into the $2,000 to $15,000 band the sales decks advertise.
Semrush's 2026 index of 126 million AI search prompts found that 45% of marketing leaders cannot accurately measure brand visibility in AI-generated answers and only 9% have tools covering all relevant metrics across platforms. The spend is real; the measurement is not.
The linear-cost problem is the same one that surfaces in multi-agent research runners when every subtask hits a frontier model. Routing and caching are what make the math close.
A Sampled In-House Pipeline Beats a Leaderboard
A reproducible internal measurement stack takes a fraction of the vendor spend and produces numbers the team can audit. The components are not novel:
- A prompt bank versioned in git, grouped by intent (brand, category, competitor, procurement)
- A scheduler that samples each prompt N times per week against ChatGPT, Claude, and Gemini using the organization's own API keys
- A parser that extracts mentioned entities and ranked positions, tolerant of citation and markdown variation
- Storage in a vector or columnar store, keyed on prompt, model, run id, and timestamp
- A weekly aggregate that reports mention rate, median position, share of voice, and run-to-run variance
Variance is the output the vendor leaderboards suppress. A weekly report that shows "brand X is named in 42% of runs with a standard deviation of 11 points" is harder to screenshot than a single rank, and more honest. Teams already running their own extraction infrastructure can plug this into an existing brand mention tracking pipeline and route LLM calls through whatever provider policy the organization uses.
Bringing in the Business Hub closes the loop. The Hub tells the team how many times ChatGPT linked to its properties over the window. The sampled pipeline tells the team how often the brand was mentioned at all, where it sat relative to competitors, and how stable the finding is. Those two numbers answer different questions and neither is a rank.
Where This Leaves Research and Marketing Teams
Teams that have not yet claimed the Business Profile should do that first and watch two weeks of Hub Insights before deciding what else is needed; the setup cost is a domain record and an admin seat. Teams already paying a third-party ChatGPT rank tracker should pull last month's invoice, divide by the number of prompts actually measured, compare the yield to what the Hub reports for free, and cancel on the next billing cycle if the delta does not justify the spend. Teams building their own measurement stack can reuse the sampled pipeline pattern above and lean on their existing extraction and routing infrastructure rather than buying a dashboard that reports a number the underlying system cannot produce consistently.
Written by
Eric LamannaEric Lamanna is VP of Business Development at Search.co, where he drives growth through enterprise partnerships, AI-driven solutions, and data-focused strategies. With a background in digital product management and leadership across technology and business development, Eric brings deep expertise in AI, automation, and cybersecurity. He excels at aligning technical innovation with market opportunities, building strategic partnerships, and scaling digital solutions to accelerate organizational growth.