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: same prompt-tracking scan, but joined server-side to your first-party session data and then to your Stripe payments. So you do not just see "we got cited" — you see "this prompt cited us, sent 12 sessions this week, and produced 2 paid trials worth $58 MRR." 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 $29/mo Pro plan, up to 100 monitored prompts scanned daily across all four major engines. 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 100 well-chosen commercial prompts cover the buying-stage query universe for most SMB SaaS and e-commerce categories. If you need a larger surface (typical for agencies tracking multiple client domains) the agency plan unlocks higher prompt counts plus workspace separation.

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 30-50 starter prompts based on your category, your existing keyword data, and the buyer-stage prompt patterns that show up consistently across the AI-citation research<Cite n={1} />. The list is fully editable: add, remove, retag intent. The system will not silently scan prompts you did not approve.

Daily on the Pro plan, across ChatGPT, Claude, Gemini, Perplexity, and Google AI Overviews. Each scan logs the engine's answer, every parsed citation, and the position of your domain within the cited source list. 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 per prompt over the chosen window. 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<Cite n={2} />.

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 now render on roughly 13-15% of US English SERPs and the trajectory is up<Cite n={3} />; ChatGPT crossed 800M weekly users and continues to grow<Cite n={4} />. 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.