Brand tracking, meet AI search

See how ChatGPT, Gemini, AI Overviews and Google AI Mode describe your brand, what that market is worth, and which of your own pages is in the way.

Overview
7d30d90d
ChatGPTGeminiAI OverviewsGoogle AI Mode
Visibility
38%+4.2
Share of voice
21%+1.8
Avg. position
2.4−0.3
Sentiment
74−2
BrandVisibilityShare of voicePosition7d
Northwind52%29%1.86
RecommaYou38%21%2.44
Latchford31%18%2.92
Openvolt24%14%3.41

The models this samples, every pass

  • ChatGPT
  • Gemini
  • AI Overviews
  • Google AI Mode

Answers are the new search results

Your buyer asks once and gets one answer. You're either in it or you aren't.

ChatGPTGeminiAI OverviewsGoogle AI Mode

“best AI visibility tracker for agencies”

For agencies managing multiple clients, the strongest options are Northwind, Recomma and Latchford.

1Northwind2Recomma3Latchford

Know what AI says, what it cites, and what it skips

best AI visibility tracker

ChatGPT · 4 samples today

Track

Every prompt, every model, every market — on a schedule

Cited sourcesLast 30 days
  • g2.com
  • searchengineland.comGap
  • reddit.com/r/SEO
  • capterra.comGap
  • producthunt.com

Attribute

See which sources feed the answer, and which ones skip you

Opportunities7 open
  • Claim Capterra listingDone
  • Pitch the SEL roundup
  • Point canonicals at the page itselfYour page
  • Add alt text to 69 images
  • Publish comparison page

Close

Earn the mentions you are missing, and fix the pages in your way

A percentage needs something to be a percentage of

Recomma prices the market behind your questions from the keywords you and your rivals are contested on — search volume, what advertisers bid, commercial intent. Visibility stops being a vanity figure and becomes a share of something: this much market, this much of it covered by a question you track, this much of that named in the answers.

$48,200/mo category value
Named in answers
41% of covered
Covered by a tracked question
$34,704
No question tracks it
$13,496

The part no question reaches is the number to act on — real demand in your category that nothing is measuring in either direction. One button turns those keywords into questions.

Reporting AI visibility for two quarters and never changing anything because of it. An action queue is where the data starts producing work.

Two queues: what to earn, and what to fix

Earned is the sources reading your category without naming you. Owned is your own pages, in the way of being read. Both ranked by what moving them is worth, then re-measured against a frozen baseline 14 days later.

Claim your Capterra listing

Reference

Cited in 5 answers across 3 tracked prompts — never mentions you

Impact3 prompts

Verified · visibility 22% → 40%

Re-measured against a baseline frozen 14 days ago

And the technical half, which needs no rival at all

Recomma reads your own site from its sitemap — not from your citations, because the pages with problems are the ones nothing has quoted yet. Which AI crawlers your robots.txt shuts out, which of fifteen markup rules each page fails, and which pages the crawler has not reached yet. None of it waits on a rival, or on a single chat being sampled.

robots.txt

  • GPTBotBlocked
  • ClaudeBotPartly blocked
  • PerplexityBotAllowed

Markup

  • Point each page's canonical at the page itself492 pages
  • Add the Open Graph type tag to your templates279 pages
  • Add alt text to the images that have none69 pages

Twenty accounts on one pool. Per-brand pricing means an agency tracks its biggest five and guesses at the rest.

Hand it to the agent doing the work

Everything above is readable over MCP, so Claude Code, Codex or Cursor can pull it into the repository where the fix happens — the answers, the gaps, the queue, the priced market, and the markup to change. Read-only, scoped to your workspace, and it sees exactly what you see.

$ claude mcp add --transport http recomma https://api.recomma.ai/mcp

Connected · 9 read-only tools

get_site_findings returns the rule, the pages that fail it, and the markup to replace — so the agent that reads it can make the change.

One prompt budget, spent where you choose

Coverage is pooled across the workspace instead of being handed out per brand, so the client who needs forty questions gets them. Brands are unlimited on Scale.

Credit pool17,500 of 20,000 used
  • Latchford Group6,200
  • Openvolt4,800
  • Cascade Health3,900
  • Meridian Legal2,600

Confidence intervals sound like a small thing. They are what stops a finance team arguing with a dashboard.

Numbers you can defend in the room

Sample size and confidence interval on every metric. Thin windows say so.

38%4.2
34%95% confidence interval43%

420 samples over 7 days

Questions we get asked

Find out what AI says about you today

Give us a domain. We read the site and propose the topics, the questions and the competitors — you keep the ones you want.