Measure, recommend, execute with approval, prove.
Pravado is one loop across three pillars. Here is what each part does today.
How often AI answers mention you, next to your competitors.
Tracked prompts
Unbranded questions your buyers ask, suggested from your industry and competitors and fully editable. Branded prompts are tagged and kept out of share metrics.
Repeated sampling, with confidence
Each prompt is sampled more than once with search turned on. Results carry the sample size and a confidence range, and a change is called movement only when the range rules out noise.
AI Share of Voice and sources
Your share of mentions against the competitors you name, with the sources each answer cited.
Refresh by plan
Starter: 25 tracked prompts, refreshed daily. Pro: 100 tracked prompts, refreshed every 6 hours. Scale: 200 tracked prompts, refreshed every 6 hours, plus an hourly watchlist of 10 prompts.
One score, always with its drivers.
EVI is built from measured inputs and shown with its three drivers: Visibility, Authority and Momentum. Scores with thin coverage are labelled provisional.
Bands: At Risk 0–40 · Emerging 41–60 · Competitive 61–80 · Dominant 81–100.
One gap, one Recommendation.
A gap is a group of prompts where competitors are cited and you are not, seen on more than one engine with enough answers to be confident. Each gap becomes one Recommendation that shows its evidence and confidence, never a promised score gain.
Pitch or pursue inclusion
Draft a pitch to a matched journalist, or work through an inclusion checklist for the review site, listicle or trade publication that is being cited.
Brief or fix a page
Generate a brief from the gap, with the sources competitors are cited from, or fix an existing page using its Citation readiness checks.
Measure the result
When the work ships, the gap is measured again and the result is reported as it is, including no measurable effect.
Your team approves every external action.
Each pillar runs in Manual or Copilot mode. Nothing is sent without an owner's or admin's approval: they can send a pitch they wrote or edited after a final confirmation, and every other pitch, including an unedited AI draft, waits in Approvals. Every action is recorded in the Activity log. Autopilot is coming soon.
The work that moves AI visibility.
PR
- Search the full journalist database on every plan, by beat, outlet, topic and location.
- Write a pitch, approve it, and send it from Pravado. Every pitch carries a one-click unsubscribe, and unsubscribes are honoured.
- Journalist email addresses are never shown or exported; an address is looked up only when you create a pitch.
- Log a placement and track its lift.
Content
- Import pages from a list of URLs or your sitemap.
- Get a Citation readiness score per page, with specific fixes.
- Create briefs from real gaps, with the sources being cited.
- Pravado prepares and approves; your CMS publishes.
Did the placement or page actually move anything?
For each placement or published page, Placement Lift tracks the prompts it should affect against a matched control set, before and after. The result is labelled proven (the engines cite the URL itself), likely (a measured lift over the controls) or no measurable effect. Until there is enough data, it says so.
Use Pravado inside Claude (via MCP).
Add Pravado as a custom connector in Claude, or another assistant that supports MCP connectors, and ask about your AI visibility, Recommendations, coverage and journalists right in the chat. It is read-only: anything that changes your work still happens in Pravado, with your approval. Your assistant's plan must allow custom connectors. Connecting an app is separate from the engines Pravado measures.
Honest by default.
- No forecasts and no promised score gains.
- No single-snapshot claims: one answer is not a trend.
- Pravado does not publish to your site, and nothing is sent without approval.
- An AI engine is named here only once it is verified live.