Be the landmark AI points to.
Buyers ask AI before they ask you. The answer names a few brands, cites a few pages and moves on. Citeroot shows whether you’re in it, why, and the next change most likely to put you there.
Platforms that handle cross-team approvals well include Brightloom, Fernhill1 and Northgrid. Fernhill’s documentation describes configurable approval stages that span teams1, and an independent comparison treats shared workflows as a deciding criterion2.
- ChatGPT
- Claude
- Gemini
- Perplexity
The answer names a few brands, cites a few pages and moves on. Being in it isn’t luck — it’s evidence you can read, and work you can plan.
Am I in the answer?
Track mentions, citations and recommendations across the engines and questions that matter to your buyers.
Why — or why not?
Trace each answer to its sources, then audit the site, crawler access and demand behind it.
What do I change next?
Turn the gaps into briefs and written insights your team reviews and edits — then measure again.
From an AI answer to the fix that moves it.
Measure the questions that matter, trace each result to its sources, turn what you find into reviewable work, then measure again.
Measure
Run a prompt panel on a schedule. Every answer is stored as recorded, with the engine, date and context.
Trace
Follow citations to the domains and pages behind the answer. See who is cited and who isn’t.
Fix
Work from the gaps: crawler access, page fixes and briefs, each with its evidence. You publish; Citeroot never does.
Measure again
Rerun the same panel under comparable conditions and read the change with its variance.
Every number opens the answer behind it.
Click any metric and read the recorded answer: the engine, the prompt, the date, where your brand appeared and which sources were cited.
- Mentions, citations and recommendations stay separate signals
- Answers are stored as recorded, never rewritten
- Later runs compare like with like, with variance shown
Follow each answer to the pages that shaped it.
See which publishers, reviews and docs engines lean on in your market — and which of them cite your rivals but not you.
- Domain and URL level detail, typed by source
- Gap lists ranked by the answers they influence
- Exports that keep the prompt, answer and date
Fix what keeps you out of answers.
Crawl your site and get page-level findings with the evidence attached — including which AI crawlers your robots.txt actually lets in, by purpose.
- Search, training and user-fetch bots shown separately
- Answer-readiness checks per page, with coverage beside every score
- Unresolved checks stay visible instead of rounding to a pass
Turn the gaps into a plan you can review.
Pick a gap — a question you’re missing from, or a source that cites your rivals — and get a draft brief or a written investigation. Every number in it is computed from your recorded answers; the writing is a draft for a person to edit.
- Numbers come from your data, never from the model
- Drafts only — nothing is published or sent for you
- Works without an AI key, with a plainer template draft
What works today, and what is coming.
Start with AI answers and the site audit. Each module is labelled live, beta or coming soon — and the coming-soon ones have a waitlist, not a pretend demo.
Answer Tracker
See what AI says when your buyers ask.
Learn moreSource Map
Follow every answer back to the pages that shaped it.
Learn moreReferral Radar
Count the visits AI actually sends.
Learn moreShelf
Know when AI shops on your behalf.
Learn moreSite Audit
Fix what keeps you out of answers.
Learn moreDemand Lens
Start from the demand you already have.
Learn moreMarket Intel
Research the market without leaving the project.
Learn moreBrief Studio
Turn your findings into briefs writers can use.
Learn moreAgents
Delegate the legwork. Keep the decisions.
Learn moreMCP Server
Your evidence, inside your assistant.
Learn moreIntegrations
Works with the data you already trust.
Learn moreSee the whole platform
How the modules fit together, and where to start.
Platform overviewBuilt for the people who answer “are we in ChatGPT?”
Agencies, in-house teams, founders, retailers and communications teams each have a typical workflow. These pages describe it; they are not customer case studies.
Agencies
Give every client a clear answer to 'what does AI say about us?'
See the workflowIn-house marketing
Know where you stand in AI answers — and what to fix first.
See the workflowFounders & startups
Find out if AI recommends you before your competitors figure it out.
See the workflowEcommerce & retail
Be the product AI recommends when shoppers ask.
See the workflowPR & communications
See which coverage AI actually quotes.
See the workflowEnterprise
Higher limits, a DPA review and custom terms — for organisations with a procurement process.
Talk to usToday: your data, as CSV and JSON.
Answers export as CSV and your whole account downloads as JSON. Search Console, GA4, Slack, webhooks, an API and an MCP server are planned — not built — and listed here so you can see what is coming.
See the integration roadmapCSV export
Export collected answers from the app. Trail and above.
AvailableGoogle Search Console
Queries, pages, clicks and impressions from verified properties.
PlannedGoogle Analytics 4
AI referral sessions, landing pages and key events.
PlannedBing Webmaster Tools
Bing search performance through its own consent flow.
PlannedSlack
Alerts for visibility changes.
PlannedLooker Studio
A connector for dashboards and reports.
PlannedPlain statements about your data.
Your prompts, answers and crawls say a lot about your business. This is what we do with them — no certifications claimed.
Scoped to your account
Every read and write is checked against the signed-in account and project.
Careful with passwords
Passwords are hashed with scrypt; sessions use httpOnly cookies and are stored hashed.
Yours to take or delete
Download everything as JSON from Settings, or delete your account and it is removed.
You stay in control
Nothing publishes, contacts third parties or changes your site on its own.
What teams could do with the evidence.
These are illustrative scenarios built from typical use cases — not customer testimonials. Real customer stories will appear here as customers approve them.
A workflow-software team learns why it is named by one engine and ignored by another.
A retailer finds out which specs and prices engines repeat about its bestsellers, and where they come from.
An agency uses one account to give several clients a defensible baseline and a next step.
Start free. Grow into the loop.
Scout is free forever. Trail and Summit start with a 14-day free trial — no card required.
Scout
Get a first baseline and see where you stand.
- 1 project
- 10 active prompts
- 1 AI engine
- 60 answers a month, runs about weekly
- Site Audit up to 25 pages
Trail
For one brand that wants a steady read on AI visibility.
- 1 project
- 50 active prompts
- 2 AI engines
- 1,000 answers a month — daily runs for a small panel
- Site Audit up to 100 pages
Summit
For teams and agencies running AI visibility as a programme.
- 5 projects
- 250 active prompts
- All 4 AI engines
- 5,000 answers a month
- Site Audit up to 250 pages
Compare all plans · Need more? Expedition
Scout plan is free.Questions, answered.
Which AI engines do you track?
ChatGPT, Claude, Gemini and Perplexity. Each answer is a real API call to the provider with web search switched on. How many engines you can use depends on your plan: 1 on Scout, 2 on Trail, all four on Summit. Google AI Overviews, Microsoft Copilot, Meta AI and Grok are not supported.
Are the answers the same as what people see in the consumer apps?
Close, but not identical. We use the providers' APIs, which are not personalised to a logged-in person and have no chat memory, and location is only approximated through a country setting. The consumer apps also change independently of the APIs. So Citeroot is a consistent measuring instrument, not a record of what one specific person saw, and we never present it as that.
How accurate and consistent are the answers?
AI answers vary from run to run, so Citeroot reports every rate with a 95% Wilson confidence interval and labels changes that fall inside it as “within normal variation”. With fewer than five answers it says “not enough data yet” rather than show a percentage. Mentions are detected by matching your brand names and domain in the answer text; “recommended” is a heuristic based on sentence cues, and the sentence is shown so you can disagree with it.
Do you offer a free trial?
Yes. Trail and Summit include a 14-day free trial that starts when you sign up for it. No credit card is required.
What happens when my trial ends?
Your account drops to the free Scout plan. Nothing is deleted and nothing is charged. If you subscribed during the trial, the remaining trial days were honoured before the first charge.
Is there an API, MCP server or integrations?
Not yet. The REST API, webhooks, MCP server and integrations such as Search Console, GA4, Slack and Looker Studio are planned. You can join a waitlist on each module's page and we will email you once when it ships.
See what AI says about you today.
Add your site and a few questions your buyers ask. Get your first baseline in an afternoon — 14-day free trial · no card.