Track whether AI assistants recommend your brand

The idea

Buying questions are moving into ChatGPT, Claude, Gemini, and Perplexity, and brands are flying blind: they cannot see whether the assistant names them, names a competitor, or hallucinates a ranking that costs them deals. AEO Engine’s $79k MRR proves budget is already flowing at this anxiety — but it sells optimization services, which requires clients to trust that content changes moved an invisible needle. Measurement is the cleaner business: a rank tracker for AI answers is a product shape marketers already understand and already pay $100+/mo for in SEO land.

The wedge is doing only measurement, honestly: multiple runs to handle nondeterminism, real response excerpts instead of a mystery score, and alerts when a competitor starts getting recommended. Renewal is driven by the same anxiety that drives rank trackers — you cannot cancel the thing that tells you when you slip.

Who pays

DTC brands, SaaS marketers, and the agencies serving them at $49-199/mo, priced like the rank trackers already in their stack. The free one-time scan is the whole funnel: nearly every marketer who runs it finds a prompt where a competitor gets named and they do not, and that screenshot sells the subscription. Distribution via marketing communities, SEO-tool comparison content, and agencies white-labeling reports for clients.

MVP scope

• Prompt suggestion, editing, and weekly multi-run tracking across three assistant APIs
• Mention parsing with alias normalization, list position, and sentiment
• Share-of-voice dashboard with weekly trends and response excerpts
• Competitor-recommendation email alerts
• Free one-time scan as the acquisition funnel
• Skip for v1: Perplexity and AI Overviews scraping, optimization recommendations, white-label PDF reports, API, multi-language prompts

Build AnswerRank, an AI-answer visibility tracker for brands.

Core flow: a marketer enters their brand, up to 5 competitors, and
their category; the app suggests 10-25 buying-intent prompts (best
project management tool for agencies, alternatives to X, is X worth
it) which the user edits. A weekly worker runs every
prompt against the OpenAI, Claude, and Gemini APIs — 3 runs per prompt
per model to smooth nondeterminism — and parses each response for
brand mentions: named at all, position in any ranked list, sentiment
of the surrounding sentence, and which competitors appear. Results
aggregate into a share-of-voice dashboard: percent of prompt-runs
naming each brand per assistant, trended weekly, with a prompt-level
drill-down showing actual response excerpts. Email alerts fire when a
competitor is newly recommended or the user's brand drops from a
prompt. Acquisition hook: a free one-time scan producing a single-run
visibility report with competitor details blurred behind signup.

Stack: Next.js App Router with TypeScript, OpenAI, Anthropic, and
Gemini APIs via a provider-abstraction layer, Postgres via Prisma for
runs and mention records, Stripe for billing.

Pages: free scan landing page, dashboard, prompt manager, competitor
view, alert settings, billing.

Non-obvious details: store full raw responses forever — the historical
archive of what assistants said becomes the moat and enables back-
testing new parse rules; normalize brand matching with alias lists
(HubSpot, Hubspot CRM) or mention counts will be garbage.

Pricing: $49/mo one brand and 25 prompts, $99/mo three brands, $199/mo
ten brands plus API access and Slack alerts.