Features
Measure, fix and prove, in one workspace
Recognly covers the whole loop: what AI engines say about you, why, what to change, who approves it and whether it worked.
- ChatGPT
- Gemini
- Google AI Overviews
- Google AI Mode
- Perplexity
- Claude
Before and after
What changes with Recognly
Three of the main features at work, from the problem to the measured result.
A wrong answer, corrected
Find the wrong fact, trace where it comes from, and work to get it corrected.
Example. Before Recognly: asked "How much does Acme Analytics cost?", an AI engine answers $40 per host per month, which contradicts the approved price of $25. After: the weekly re-check finds the answer matches the approved fact, $25 per host per month.
Appearance rate, against the holdout
Measure the questions you work on next to questions left untouched.
Example. Before Recognly: in week 1 the brand appears in 18% of answers to worked questions and 17% to holdout questions. After 12 weeks: worked questions 34% (95% CI 28 to 40), holdout 19%, a lift of 14 points (95% CI 6 to 22).
A draft through review
Replies arrive drafted and checked; your team approves and your expert posts.
Example. Before Recognly: a buyer's question in a public community gets answers from competitors only. After: a reply is drafted from approved facts with a disclosure, passes the compliance checks, is approved by a person on the team, and the named expert posts it by hand.
Answer tracking
How six AI engines answer your buyer questions, measured on a schedule with confidence intervals.
Add the questions your buyers actually ask. The platform samples answers from each engine on a schedule and reports how often your brand appears, which competitors show up and which sources get cited.
- Appearance rates always carry a 95% confidence interval and the number of answers behind them
- Answers captured from consumer interfaces and answers from official APIs are reported as separate panels, never blended
- Every answer is stored so you can read exactly what the engine said
Appearance rate by engine, last 4 weeks
Consumer answers (UI capture)
- ChatGPT31% (24 to 38)
- Gemini22% (16 to 29)
- Google AI Overviews18% (12 to 25)
- Google AI Mode26% (19 to 33)
Model behaviour (API)
- Perplexity35% (28 to 42)
- Claude20% (14 to 27)
Dots show the rate, lines the 95% confidence interval. 120 answers per engine. The two panels are reported separately and never blended into one number.
Company Brain
The approved facts about your company that every draft is checked against.
Facts are extracted from your own pages with the passage they came from, then approved by your team. When two of your pages disagree, the conflict goes to the person who owns that decision instead of being guessed.
- Each fact links to its source passage and records who approved it
- Conflicts such as two different prices are flagged for an owner
- Answers that contradict an approved fact are logged as wrong-fact issues
Company Brain
- Approved
SOC 2 Type II report available on request
Source: /security
- Needs owner
Plans start at $49 or $59 per user per month
Source: /pricing and /plans disagree
- Approved
Data can be hosted in EU or US regions
Source: /docs/regions
Site audit
Finds the technical issues that stop AI engines from reading your site correctly.
The audit crawls your site, checks your entity signals and looks at what third-party pages say about you. Technical readiness and external visibility are reported as two separate measurements, and failing checks become tasks with a clear fix.
- Canonical faults, orphan pages, hidden text and structured data
- Entity checks that engines describe the right company
- Failing checks flow straight into the action plan
Technical readiness
- Canonical tags consistent: passed
- Organization schema matches the page: passed
- FAQ answer hidden until clicked on /pricing: failed
- Sitemap lists every product page: passed
Opportunities
Public conversations where your buyers ask the questions you track, scored for relevance.
Discovery reads public threads through official APIs and feeds, scores them against your buyer questions and queues the best ones. You can also import a list or save a thread by hand.
- Hacker News, GitHub, Stack Exchange, X and RSS feeds, plus imports and saved links
- Pages that AI engines cite when you are missing become outreach targets
- One client per conversation, enforced across the platform
Opportunities queue
- 86Brief only
Hacker News
Ask HN: How do you keep audit logs across many Postgres clusters?
- 74Reply
GitHub Discussions
Exporting audit events to S3 on a schedule?
- 71Outreach
Cited page
Blog roundup of audit logging tools that engines cite without Acme
Drafts with compliance checks
Fixes for your site and replies for your named experts, checked before anyone approves them.
Drafts are grounded in approved facts. For community replies and outreach, the disclosure sentence is inserted automatically and cannot be edited out. A person on your team approves every draft, and a named expert posts it by hand from their own account.
- Disclosure, grounding, vote-request and similarity checks run on every edit
- Website fixes go out as pull requests or CMS drafts for your team to publish
- The software never posts to third-party platforms
Reports
Weekly and baseline reports your team and your leadership can read in five minutes.
Reports summarise appearance rates with their intervals, lift against the holdout, competitor movement, cited sources and the work shipped. Download them, email them on a schedule or share a read-only link.
- HTML, CSV and JSON exports
- Scheduled weekly or monthly emails
- Expiring share links for people outside your workspace
Weekly report
- Appearance rate
- 34%Consumer answers (UI capture). 95% CI 28 to 40, 312 answers
- Appearance rate
- 29%Model behaviour (API). 95% CI 23 to 36, 180 answers
- Lift vs holdout
- +14 ptsConsumer answers (UI capture). 95% CI +6 to +22
- Drafts approved
- 64 posted, 2 pull requests
Alerts
A heads-up when visibility drops, a new competitor appears or an engine states a wrong fact.
Alerts run after each measurement and are de-duplicated per week, so your team hears about real changes once.
- Visibility drops, new competitors, wrong facts and lost citations
- In-app, email and Slack delivery
- Signed webhooks for your own systems
Alerts
Visibility drop on Perplexity
35% (95% CI 27 to 43) to 24% (17 to 31) week over week, 140 answers each week
New competitor named
Globex Audit appeared in 9 answers this week
Wrong fact
An answer says Acme has no EU region
In-app, email, Slack and webhooks
Action plan
One prioritised list of what to do next, built from every signal in the workspace.
Failing audit checks, wrong facts, strong opportunities, facts waiting for an owner and questions where competitors appear without you become tasks, ordered by expected impact.
- Assign tasks, set due dates and comment with mentions
- Tasks close themselves when the underlying issue is resolved
- A personal view of everything assigned to you
Action plan
- Assigned to JL
Show the FAQ answer on /pricing without a click
Audit · High impact
- Assigned to MR
Confirm the current starting price
Company Brain · High impact
- Assigned to PS
Write a reply to the Ask HN thread yourself
Opportunity · Medium impact
- Assigned to JL
Publish a retention comparison page
Visibility gap · Medium impact
AI-ready assets
llms.txt, Organization and FAQ schema and a fact sheet, generated from approved facts only.
Assets are built deterministically from your brand settings and approved Company Brain facts. Nothing is invented, and each file notes that a person should review it before publishing.
- llms.txt and llms-full.txt
- organization.jsonld and faq.jsonld
- A plain-language fact sheet
AI-ready assets
- llms.txt
- llms-full.txt
- organization.jsonld
- faq.jsonld
- fact-sheet.md
# Acme Analytics > Audit logging for Postgres teams. ## Product - [Pricing](https://acme.example/pricing) - [Security](https://acme.example/security) ## Docs - [Retention](https://acme.example/docs/retention)
Built from approved facts only. Review before publishing.
Industry benchmarks
How your appearance and citation rates compare with anonymous peers in your industry.
Benchmarks are computed from opted-in brands per industry and engine. Nothing is shown for a cohort smaller than five brands, and any illustrative reference data is labelled as such.
- 25th, 50th and 75th percentiles per engine
- Your position within the cohort and the cohort size
- Organisations can opt out of contributing
Benchmark: appearance rate, ChatGPT
- 25th
- 14%
- Median
- 21%
- 75th
- 29%
- You
- 31%23 to 40
Your rate: 31% (95% CI 23 to 40, 120 answers), consumer answers panel (UI capture). Cohort: 14 anonymous opted-in brands in the same industry. Benchmarks are hidden below 5 brands.
CRM attribution
Leads and pipeline that arrived from AI engines, from HubSpot or Salesforce.
Connect your CRM to see leads and deal value whose first touch came from an AI engine, alongside the visibility work that preceded them.
- HubSpot and Salesforce connections over OAuth
- Leads grouped by channel and by AI engine over time
- Email addresses are stored hashed, never in plain text
Leads from AI referrals, last 90 days
AI referral leads
24
Pipeline value
$86,400
- ChatGPT11
- Perplexity7
- Gemini4
- Claude2
From HubSpot. Emails stored hashed.
Agency portal and white-label
Run many client brands from one workspace and give each client a branded view.
Agencies manage every client brand in one portfolio, invite clients as guests who see only their own brands, and send reports under their own name and domain.
- Portfolio view with visibility, issues and pending drafts per brand
- Client guests who can review and comment on their brands
- White-label reports on your own verified report domain
Client portfolio
Acme Analytics
2 drafts to review
34%
95% CI 28 to 40
312 answers
Globex Cloud
Nothing waiting
21%
95% CI 16 to 27
264 answers
Initech HR
4 drafts to review
27%
95% CI 21 to 33
288 answers
Rates are for the consumer answers panel (UI capture). Clients sign in as guests and see only their own brands.
Enterprise SSO
Sign-in through your identity provider with OpenID Connect.
Connect Okta, Microsoft Entra ID, Google Workspace, Auth0, OneLogin, JumpCloud or any OpenID Connect provider. Verify your email domains and optionally require SSO for members.
- Just-in-time provisioning with a default role
- Domain verification by DNS record
- Optional enforcement, with owner break-glass access
Single sign-on (OpenID Connect)
- Issuer
- https://login.acme.example
- Client ID
- 0oa8f2k1x9
- Default role
- Member
Require SSO for acme.example
Domain verified
On its own
What runs automatically
You do not have to start these. Times are UTC unless a line says otherwise. A paused brand, or an organization whose plan is not active, does not run.
Every week
- Your weekly run, on the day and hour you set (Monday 06:00 in your time zone unless you change it). Each brand starts a little after the hour, so all brands do not start at once. Plans with scheduled runs, and the trial while it lasts.
- In that run: your buyer questions asked to each AI engine, the analysis, the due re-checks, the site crawl and audit, the brand safety scan, drafted fixes for your team to review (plans with drafts) and the weekly report.
- Your CRM sync and proof-of-impact estimate, right before the report, when a CRM is connected or an impact plan is due.
- The report email, as soon as the weekly report is ready. It goes to team members who keep report emails on and to the addresses you add (plans with report sharing). Weekly is the default; you can choose monthly or off.
- Your digest email, on Monday from 08:00 in your own time zone, if you keep Digests on. You can switch it to daily.
Every day
- Wrong-fact re-checks that are due, looked at every 6 hours (00:27, 06:27, 12:27 and 18:27 UTC).
- Compliance re-checks that are due, started every day at 06:22 UTC.
- Your daily digest, from 08:00 in your own time zone, if you chose daily digests.
Every month
- Deep research, on the 3rd at 04:20 UTC, if your plan includes it and you turned it on for the brand. A month that already has its reports is not run again.
- The report email, on the 1st to the 3rd from 09:00 in the brand's time zone, if you chose monthly report emails.
Once, during the trial
- One deep-research run per brand, started on its own once the trial reaches day 7 (checked daily at 04:33 UTC), unless you switched deep research off.
- A reminder email to the owners and admins when the first charge is 3 days away.
Coming soon
From AI visibility to real leads
We are working on tying what AI engines say about you to the customers it brings in.
Try the whole loop on your own brand
Start a 14-day free trial on one brand with all six engines, plus an example brand to explore.