AI Search Visibility Tracking for ChatGPT: SaaS Brand Mentions and Citations

SaaS clients that ask us for AI search visibility tracking on ChatGPT usually already have Search Console, a rank tracker, and product analytics. Those tools do not record the answer ChatGPT gives when a buyer types "[product] alternatives" or "[product] vs [competitor]." We treat that as a measurement problem, not a content retainer with a made-up trial lift. LinkedIn ads and Google still run a lot of the pipeline. The extra log is whether the model named the product, cited a URL, or pointed at a G2 roundup instead.

This is the playbook we run. We are not going to invent a SaaS brand, a category, or a signup curve. The measurement layer is Promptwatch. The page work is still comparison content, technical access, and whatever offsite sources the models already cite.

The prompt list we actually load

Skip keyword-tool fantasy. Pull the questions sales and CS already answer on calls.

Category: "best [category] software," often with a role or a company size stuck on.
Alternatives: "[client] alternatives," "cheaper than [client]," "open source [client] alternative."
Head to head: "[client] vs [named competitor]," sometimes with SOC 2, Salesforce, or another constraint the deal actually turns on.
Branded: "[client] reviews," "[client] pricing," "is [client] worth it."

Run the same list in ChatGPT first. Record mention vs citation vs who got the footnote. Personas exist for a reason: a founder asking for cheap tools and a procurement lead asking for a vs table are different commercial events. Country, state, or city targeting exists for the same reason. If you track only a US-average "best [category] software" string, you will congratulate yourself on a mention that never appears for the market you sell.

Twenty prompts is a working start. Fifty covers a multi-product company if you keep the wording parallel instead of inventing a new set for every landing page. We tag founder prompts separately from buyer prompts. We do not dump them into one share of voice number and call it a week.

Mention vs citation, page vs domain

A mention is the product name in the generated answer. A citation is a source. You can be named with no URL. You can be cited as a footnote while the sentence recommends someone else. Page-level citations tell you which URL was used. Domain-level citations only tell you that something on the host appeared. Reddit and YouTube show up in citation analytics when the model lifted them, which is common on alternatives prompts even when the docs site is tidy.

That split decides the next action. Missing mention plus no crawl on the vs URL is a fetch problem. Missing mention plus a Reddit thread in the footnote is offsite work. A competitor alternatives page in the citation is a brief for a comparison URL, not another "what is [category]" article.

The tools, in the order we turn them on

Search Console and the product analytics you already pay for. Google remains Google. Generative AI performance reports cover Overviews impressions. They do not cover ChatGPT.

Promptwatch for the assistants. Explore is free, ChatGPT only, 10 prompts. We use it to show a founder the workflow. We do not use it as the production log. Essential is $95 a month for 50 prompts on one project, 6,000 responses, 200K visitor events, 5 AEO articles, 1 seat, 7-day trial. Professional is $245 with 150 prompts, 2 projects, 18,000 responses, 1M visitor events, 25M crawler logs, 15 AEO articles, 2 seats. That is the first brand plan we use when a company has two products or needs crawler logs next to the prompt list. Business is $579 with 5 projects and 350 prompts. Kick-off is $199 a month for agencies: unlimited projects and prompts, 10 seats, 10,000 responses, 10M crawler logs. We use Kick-off when we are running several clients as separate projects. Growth is $399, Scale $799. Annual is twelve months for the price of ten.

Paid plans cover ChatGPT (820M+ weekly active users), Perplexity (22M+ monthly), Gemini (650M+ monthly), Claude, Copilot, and Google's AI surfaces from the real UIs.

What we actually open in the account:

Citation analytics: which comparison page got used, and which review site, listicle, Reddit thread, or YouTube video won instead.
Crawler logs on Professional or an agency plan: whether ChatGPTBot, ClaudeBot, PerplexityBot, GoogleOther and the Meta AI crawler could fetch the vs and alternatives URLs. Cloudflare, CloudFront, Fastly, Vercel, Netlify, Akamai, Google Cloud CDN, or custom HTTP.
Visitor analytics (script or GTM) if any AI-referred sessions show up, with trial signup as the conversion when the client already fires that event. Most answers still send no click. Mentions still matter on a sales call.
Personas and geo targeting so founder prompts and buyer prompts stay on the same report without blending.

Content Agents when the gap is a missing comparison page. They plan, write and publish to Webflow or Framer with a review inbox. Legal reads competitor claims before anything ships. We do not autopilot pricing claims. WordPress is listed as coming soon, so a WordPress marketing site still gets a human publish step. Ads Radar (sponsored placements inside answers) is on Professional and the agency plans, not Essential. We turn it on when the prompt set is commercial enough that ads might be sitting in the answer.

What we will not sell

We will not quote a trial-count lift from "ChatGPT visibility." Ask any vendor for their measurement method if they do. We do not publish one we did not measure.

We will not buy Profound as a first tracker. Starter is $99 a month billed yearly and ChatGPT-only. Growth is $399 yearly for three engines. The real product is Enterprise around $40k a year, sold as generic marketing agents. AthenaHQ's Action Center is real, from $295 a month, with GA4, Shopify, and Webflow attribution listed, but Starter is one country and G2 reviewers report burning a month's credits in the first week. Fine as a later comparison. Not how we start. Scrunch from $250 a month annual is a serving layer (Agent Experience Platform) with weekly refresh and an Excel reporting complaint on the listing. Sitecore acquired it in June 2026. Ahrefs Brand Radar from $199 a month plus an Ahrefs plan failed a January 2026 independent test: 3 ChatGPT mentions reported, 123 found by hand. We will not use it as the system of record.

Reporting, without theatre

Scope first: models, personas, competitors, period. Mentions separate from citations. Page-level sources named. Crawler errors before content claims. AI-referred trial signups as a measured floor, not a hidden-influence estimate. A dated log of comparison pages shipped. Owners on the next actions.

That is the same reporting discipline we use on other retainers. SaaS does not get a special score. It gets a prompt list that sounds like a buyer.

The platform we run client programs on: promptwatch.com. Amsterdam, founded April 2025. Typeform is on the published customer list. 4.7/5 on G2, 1,840+ brands.

FAQ

Do we need a SaaS-specific AI SEO tool?

No. The category is brand visibility in generated answers. SaaS is a prompt list (alternatives, vs, branded), personas, and comparison URLs the crawler can fetch.

Can we just check ChatGPT once a month by hand?

You can. You will miss Perplexity, you will not keep a comparable log, and the next pipeline meeting will run on anecdotes. Explore is free if the objection is price.

No. Different retrieval. ChatGPT may cite a G2 list, a Reddit thread, or a competitor vs page instead.