Track prompts. Run the engines. Read your SeenRate.

1. Tell us about your business

A short description, your keywords, and your website. That is enough — Frontpagely works for any business, and generates the buyer prompts that matter for your market from your own inputs. Industry presets exist if you want a head start, but nothing is hard-coded to one vertical.

2. We run the engines on a schedule

Your prompt set runs against ChatGPT, Claude, Perplexity, and Gemini on your plan’s cadence. Each answer is checked: were you named, at what position, who was named instead, and which sources the engine cited. Because AI answers vary, we sample and aggregate rather than trusting a single response — and low-confidence results are labeled directional, never dressed up as certainty.

3. Read one blended SeenRate

Answer visibility is the biggest component of your 0–100 SeenRate, blended with your SEO standing (authority and rankings, which the engines lean on) and an automatic on-page site grade (schema, metadata, performance, security). One number up top; per-engine and per-prompt evidence underneath.

4. Work the fix plan — and prove the lift

Every run produces a ranked fix plan: schema to add, profiles to claim and reconcile, content gaps tied to the prompts you are invisible for. On paid plans, completed fixes are verified against measured signals — the score moves when the fix actually lands, not when a checkbox is clicked. That is the loop: measured, fixed, proven.

The methodology — how the score is computed

SeenRate is a 0–100 score built from three measured families, each weighted by how much it drives whether AI assistants recommend you. The weights are public and versioned:

Confidence gating, in plain words: every family carries a confidence value based on how fresh and how thick its data is. A family below the confidence bar is left out of the blend entirely — the remaining weights stretch to cover it — rather than counted as a fake zero. And when the overall sample is too thin to trust, the score itself is labeled directional. We would rather show you an honest “low confidence” than a confident-looking number we can’t stand behind.

The verified-lift bonus: fixes you complete can add a small, capped bonus (at most 6 points) on top of the measured blend — but only once a fix is verified against a measured signal. The bonus renders as its own labeled band, never silently folded into the measured score, and it expires when the next full measurement supersedes it.

Versioned, never silently rebased: when the formula changes, the version changes — scores are never silently rebased. Every score you have ever seen is stamped with the methodology version that produced it, so your trend stays comparable over time instead of quietly shifting under you.

What the labels on your fixes mean

Not all “done” is equal, and the product never pretends otherwise. Every completed fix carries one of three labels:

That ladder is the product’s honesty model in one table: a click never sets “verified” — evidence does.