Prompt tracking that ends
at the Stripe payment.

Prompt tracking is the practice of monitoring a fixed set of buyer-stage prompts on AI engines (ChatGPT, Claude, Gemini, Perplexity) over time, logging whether each prompt cites your domain, your competitors, or neither. It is the AI-search equivalent of rank tracking in classic SEO — but where rank tracking measures a single position in a SERP, prompt tracking measures presence inside a generated answer. The reason it matters is that AI answers replace SERPs for an increasing share of buying-stage queries; if you cannot see which prompts cite you, you cannot tell why your AI traffic shifted week over week. The honest version of prompt tracking always ends at revenue, because a tracked prompt is only useful if the citation it monitors eventually produces a paying customer.

SEOcrawl runs a clean per-LLM prompt-tracking product across ChatGPT, Claude, Gemini, Perplexity, and Copilot, and they report citation appearance, source share, and trend lines. What they stop short of is the join to your actual traffic and revenue. Their dashboard ends at "you were cited in N answers." Attrifast picks up where they stop: the same per-query scan gives you a cited/uncited matrix for every monitored prompt across ChatGPT, Claude, Gemini, and Perplexity, and next to it the revenue attributed to each AI engine — sessions, paid trials, and Stripe revenue joined at the engine level. So you do not just see "we got cited" — you see which prompts cite you and which engines actually produce paying customers. That is the loop SEOcrawl leaves open. Pick the better tool for your job: if you only need citation monitoring, SEOcrawl is fine; if you need to defend a marketing budget, you need the revenue join.

On the $49/mo Pro plan, up to 30 prompts monitored, rescanned every 10 days across ChatGPT, Claude, Gemini, and Perplexity. The $9.99/mo Starter plan tracks up to 10 prompts, rescanned every 30 days. That is intentionally fewer than the high-end enterprise plans of competitors — the goal is depth of measurement per prompt, not vanity count. In practice 30 well-chosen commercial prompts cover the core buying-stage query universe for most SMB SaaS and e-commerce categories.

Both. You bring whatever prompts you have — from a sales call, from a customer survey, from your GSC top-query report — and we will scan them. During onboarding we also suggest starter prompts (up to the prompt limit on your plan) based on your category, your existing keyword data, and the buyer-stage prompt patterns that show up consistently across the AI-citation research. The list is fully editable: add, remove, retag intent. The system will not silently scan prompts you did not approve.

Scheduled scans run every 10 days on Pro and every 30 days on Starter, and you stay in control: Attrifast runs the scheduled scan and you can also trigger a rescan from the dashboard, at most once every 7 days on Pro (14 days on Starter), across ChatGPT, Claude, Gemini, and Perplexity. Each scan logs the engine's answer, every parsed citation, and whether your domain was cited on each prompt. Trend charts are then computed against the full historical scan record, so you can see when a model retrain dropped you out of an answer, when a fresh competitor displaced you, or when an answer to a specific prompt started linking to a different domain on your site.

The chain is mechanical. (1) The prompt-tracking scan logs that engine E cited your domain on prompt P at time T. (2) When a visitor lands from engine E (chatgpt.com, perplexity.ai, etc.), the Attrifast tracker captures the referrer server-side and writes a first-party session id, including the engine source. (3) When that session later completes a Stripe checkout, the payment_intent.succeeded webhook fires and the payment is joined to the session and back through the session to the engine. (4) Reporting aggregates payments per engine over the chosen window — revenue is attributed at the engine level, shown alongside the per-prompt citation matrix that tells you which prompts each engine cites you on. The chain breaks if you stop one step short, which is exactly what GA4 does — it strips most AI engine referrers into the Direct bucket.

Yes for the major Latin-script languages (Spanish, French, German, Portuguese, Italian, Dutch) and for the engines that return cited URLs on those locales. Citation parsing is language-agnostic when the engine emits a structured citation array; for engines that only emit bare brand mentions, accuracy depends on entity disambiguation in the target language and is materially weaker for CJK scripts in our current build. We are honest about this on the pricing page rather than waving the flag.

Not yet, but it is a separate scoreboard that matters more every quarter. AI Overviews render on roughly a fifth to a quarter of tracked SERPs (Ahrefs measured 20.5% of 146.1M SERPs in September 2025; Conductor 25.11% of 21.9M searches that autumn); ChatGPT crossed 800M weekly users and continues to grow. Classic rank tracking still pays because Google organic still drives the majority of trackable search traffic for most sites. Prompt tracking pays because the slice it measures is the slice growing fastest, and because the citation event is invisible to your rank tracker. Most teams I work with now run both, weighted by how much of their actual revenue already comes from AI engines.