For agencies serving B2B SaaS clients
PromptMentions records how ChatGPT, Perplexity and Claude answer the questions buyers ask, keeps every answer as evidence, and turns what it finds into tasks, reviewed content and client reports.
7-day trial · one client · no card needed. Live collection and connected services need setup once you are in.
Evidence first
AI-visibility tools are easy to fake with a score. PromptMentions is built the other way round: it keeps the raw answer, says exactly how it was collected, and refuses to fill gaps.
Provider, interface, model, prompt revision, locale, time and sample size sit beside every rate. Open any figure and read the answer behind it.
Illustrative label, not a measurement:
Failed and disconnected samples are excluded from the denominator and counted separately. A missing observation is shown as missing, never as zero. There is no composite score.
Briefs rest on approved facts. A draft becomes content only when someone approves an exact revision. Publishing goes to a CMS draft with a receipt you can inspect.
How it works
Monitor, find the gap, fix it, prove it. Each step keeps the evidence of the one before, so nobody rebuilds the story for the monthly update.
Save the questions a buyer would ask an AI assistant, then collect one answer per prompt and provider through official APIs. Every answer is kept whole as evidence.
See which sources AI answers cite, where competitors appear instead of the client, what the answers get wrong, and whether AI crawlers can even read the site.
Turn an observation into a task with its evidence attached, write a source-backed brief, draft, review and approve an exact revision. Nothing is published without a person saying so.
Give the client a report built from recorded observations and completed work for a date range, printable and shareable with the right permissions.
What you get
Each client is its own project with its own prompts, evidence, tasks, content and report. Switch clients from anywhere; nothing leaks between them.
Owners run billing and connections, editors do the work, viewers read reports. Invitations are tied to a verified email.
The overview tells you what to do next and what needs attention: an unverified website, a failed run, ageing results or a decision waiting on you.
Samples and drafts are reserved before anything is sent and settled once. Usage and budget state are always on screen; there is no unlimited plan.
A client report for any date range: observations, changes and completed work, with CSV export and permission-checked sharing.
Workspace keys with scopes and client allow-lists expose the same data to your dashboards and AI agents. Developer docs
Pricing
No invented tiers here: this is what exists today.
Free7 days · no card needed
Per workspaceGBP, monthly, published when billing opens
Allowances, trial length and plan limits are enforced in the product and shown under Billing & usage. Prices are not listed until they are live, so that this page never promises something the product does not do.
Questions
No, and PromptMentions never claims it is. Observations come from official provider APIs and carry the provider, model, prompt revision, locale, time and sample size. They are evidence of how a model answered a given prompt at a given time, not a copy of a consumer screen or a search ranking.
OpenAI, Perplexity and Claude through their official APIs for collected answers; Bing Webmaster AI-performance exports can be imported and are kept separate. Other engines are listed as not connected until they are, rather than estimated.
No. Drafts stay drafts until an authorised person approves an exact revision. WordPress delivery sends a draft to a connected site and records the receipt; publishing live is a separate, confirmed step.
They are excluded from every rate and said to be. A run that collected nothing shows nothing. There is no composite score and no invented trend.
Each client is a project inside your agency workspace. Every read, write, background job and evidence download is checked server-side against the workspace, the project, the member’s role and the live session.
A new workspace starts with a 7-day trial that includes one client and a monthly allowance of answer samples and AI drafts, with no card required. Paid plans are priced per workspace in GBP and published when billing opens; nothing is charged without a subscription you choose.
Create an account, name your workspace and add a client website. Prepare prompts and briefs straight away; connect collection when you are ready.