How can a B2B SaaS track its visibility in ChatGPT? How we set it up
How we set up ChatGPT visibility tracking for B2B SaaS clients: the prompt types that matter, the pages models cite, and tying AI referrals to demos and signups.
SaaS clients tend to understand quickly why ChatGPT matters. Their buyers already use it to build shortlists, and a vendor missing from the shortlist rarely gets a chance to pitch. What they ask us is how to track it properly. Here is the setup we use, and the reasons behind each part.
Start with how software gets bought
Software buyers ask in patterns, and the prompt set should follow. We sort prompts into groups:
- Category questions: "best help desk software for startups."
- Use-case questions: "how to automate vendor onboarding."
- Comparison questions: "Tool A vs Tool B."
- Alternatives questions: "alternatives to Tool B," where many shortlists begin.
- Pricing and integration questions: "does Tool A integrate with HubSpot," "Tool A pricing."
- Brand questions: "is Tool A secure enough for enterprise."
We tag by topic and by persona, because a head of finance and an engineer ask differently. Promptwatch's persona stacking lets us run one prompt through several buyer roles, which keeps the list manageable.
The pages that matter most
For SaaS, the pages that answer the pricing, integration and comparison prompts are the ones to watch: pricing, integrations, security, docs and your own comparison and alternatives pages. Models answer pricing questions from pricing pages, and if a model cannot fetch yours, it uses someone else's version. We check crawler access on these pages first.
Our client programs run on Promptwatch. Agent Analytics shows real-time crawler logs, verified by IP, for OAI-SearchBot, ChatGPT-User and the other real AI bots, with a citation rate per page. If the pricing page is fetched but never cited, we look at content. If it is not fetched at all, we fix access. That split saves a lot of guessing.
Measuring the answer
For ChatGPT, we track Visibility Score, Share of Voice, Position in Answer and Sentiment on a daily panel, against three to five named competitors. Page Tracker lets us follow a competitor's alternatives page, which is often the most cited page in the whole set. Citation analytics shows the third-party sources, such as review sites and community threads, that models lean on for software questions.
Attribution is where SaaS gets interesting
Referral traffic from chatgpt.com appears in standard analytics, but it only captures clicks. Visitor analytics in Promptwatch tags AI-referred sessions and conversions, so we can follow a visit to a demo request or a signup. Search Insights adds Google Search Console and Bing Webmaster traffic next to AI crawls and citations, so the client sees classic and AI performance together. We install it with the lightweight script, a GTM template or the WordPress plugin, depending on the stack.
From findings to work
The weekly review follows a fixed shape: lost prompts, competitor moves, cited pages, crawl errors, AI-referred conversions. The Action Board collects tasks. For clients on Business plans, the Content Agent can plan and publish against gaps through WordPress, Webflow or Framer, and we keep a review step in front of everything.
Setting expectations
We tell SaaS clients that AI visits will likely be a small share of traffic at first, and that the value is in learning where they win and lose early. Nobody can guarantee a recommendation, since answers vary by run and user. We track trends across a fixed prompt set, and we do not quote numbers we cannot show.
Other tools can cover parts of this. AthenaHQ has an action queue, Scrunch AI has audits, Profound sells enterprise agents, and Otterly.AI and Peec AI are quick trackers. We settled on one platform for the crawler logs and conversion attribution, since SaaS clients ask for pipeline evidence early.