Marketing & GrowthFree tool
AI Search Visibility Report for Your Brand
A report on how often AI assistants name you when buyers ask about your category, which sites they cite, and how to get onto each.
Free to start · Opens in your River workspace, ready to edit and share
How to track brand mentions in AI search results
To track brand mentions in AI search results, put your buyers' questions to each AI assistant several times and log every answer. Record whether your brand was named and which pages the answer cited, then report each as a share of runs. Rank the sites cited most often, since those are the sources the answers draw on. Before you start, check that each assistant's search crawler can reach your site.
Keep those two counts apart. Being named shows how assistants see you. Citations show what to do next: each is a page someone publishes, which you can improve, earn a place on, or mark as a dead end.
Check that assistants can reach you
Start with access: an assistant that can't crawl your pages won't cite them. OpenAI says sites that block OAI-SearchBot will not be shown in ChatGPT search answers1. Its training crawler, GPTBot, is a separate setting, so you can opt out of training and still appear. A rule meant only for training still costs you citations if it blocks every bot.
At Google, Googlebot controls crawling for Search, AI Overviews and AI Mode included, while Google-Extended covers training and grounding in some other Google systems. To appear as a link in those answers, a page must be indexed and eligible to show a snippet2. Since August 31, 2026, every site has also had a Search Console switch that can opt it out of both3.
Five checks before you measure
| Check | Where to look | What a block costs |
|---|---|---|
| ChatGPT search | OAI-SearchBot in robots.txt | Not shown in ChatGPT search answers |
| Google crawling | Googlebot in robots.txt | No links in AI Overviews or AI Mode |
| Snippet controls | nosnippet, max-snippet and data-nosnippet on key pages | nosnippet rules a page out as a link; the others limit what Google shows and uses |
| Search Console | The generative AI switch in Settings | No traffic or impressions from AI Overviews or AI Mode |
| Firewall and CDN | Bot rules at your host or CDN | Crawlers turned away even when robots.txt allows them |
For the rest of crawling and indexing, work through a technical SEO audit.
Fix access before judging content
A zero where an assistant can't reach you says nothing about your pages. Fix the block, give crawlers time to return (OpenAI estimates a day for ChatGPT), then measure again.
Write the questions buyers ask
List the questions a buyer asks before they know your name, in their words: the problem, the kind of solution, comparisons and price. Sales calls, support tickets and Search Console queries are good sources, and a keyword clustering template groups them by need. A freight audit firm's set might start like this:
- Problem: "How do I catch billing errors on freight invoices?"
- Solution: "What does freight audit software do?"
- Comparison: "Freight audit service or our TMS's built-in checks?"
- Price: "How much does freight bill auditing cost?"
- Shortlist: "Best freight audit companies for mid-size shippers"
Wording matters less than covering each need. In a study published in January 2026, SparkToro, an audience research company, found that people almost never phrase the same need the same way4. Yet the assistants' answers still named a fairly consistent set of brands. SparkToro ran the study with Patrick O'Donnell of Gumshoe, an AI tracking startup.
Aim for 20 to 30 questions, one or two phrasings each, and freeze the set before the first run. Change it between rounds and you can't tell a real shift from a new question.
Run each question several times
Expect answers to change from run to run. In the same SparkToro study, 600 volunteers put 12 prompts to ChatGPT, Claude and Google's AI a combined 2,961 times. The study found less than a 1 in 100 chance4 that ChatGPT or Google's AI, asked 100 times, would give the same list of brands twice.
SparkToro concluded that a brand's share of runs, across many prompts, is a reasonable measure, but its position in the list is not. Never judge from one screenshot.
Use a clean session
Open a private window, logged out where the assistant allows it, so your history doesn't shape the answers.
Fix the run count
Ask each question the same number of times in every assistant, such as five. At five runs, one answer moves a question's share by 20 points.
Log one row per answer
Date, assistant, question and run number, whether you were named, and every link the answer cited.
Open every cited link
Log the page that loads. Testing news searches in early 2025, Columbia's Tow Center found that more than half of Gemini's and Grok 3's responses cited fabricated or broken links5.
Count brand mentions and citations separately
Being named and being cited are different results with different fixes. An assistant can name you from what it already knows while citing a rival's page. Report the two side by side, never as one score.
Measure the answers themselves, because clicks won't show them. Pew Research Center tracked the browsing of 900 US adults in March 2025. When Google showed them an AI summary, they clicked a link inside it on just 1% of those visits6.
Calderwood Audit, a freight audit firm: 20 questions, 5 runs each in ChatGPT, Google AI Mode and Perplexity, 300 answers.
Named in 54 of 300 answers (18%). Own site cited in 12 (4%).
ChatGPT named Calderwood in 9 of 100 runs and cited its site in none: robots.txt blocks OAI-SearchBot.
Never named on 4 of the 20 questions, including "How much does freight bill auditing cost?"
- Access first. The ChatGPT zero is a blocked crawler, not weak pages.
- Named isn't cited. The assistants know the name but seldom use Calderwood's pages as evidence.
- Zeros are the work list. Each never-named question shows which sites were cited instead.
Turn cited sites into a work list
Count citations by domain across all runs and rank the domains by how many runs cited them. A few sites recur: in Pew's sample, Wikipedia, YouTube and Reddit were the sources Google's AI summaries cited most6, 15% of their sources between them.
Cited sites and the route onto each
| Cited site | Route onto it |
|---|---|
| Your own site | Answer the question plainly, and keep crawlers allowed |
| Review sites | Ask real customers to review you there |
| Trade press and blogs | Pitch a story, original data or a contributed article |
| Reddit and forums | Answer under your real name, saying where you work |
| "Best of" roundups | Ask the author to consider you, with evidence they can check |
| Directories and associations | Get listed, or join |
| Rivals' pages and encyclopedias | No fair route. Read a rival's claims with a competitor teardown, then move on |
Then list every question where you were never named, with the sites cited instead: those pages could have named you. Often the quickest fix is on your own site, a plain answer or an FAQ section built from real questions.
Skip the shortcuts
Google says you don't need new machine-readable files, AI text files or markup2 to appear in its AI features. Never post as a fake customer or pay for positive reviews.
Rerun the same set each quarter
Keep the questions, run count and assistants the same each quarter, and compare shares, not single answers. Search Console now also reports your impressions in AI Overviews and AI Mode, and which pages appeared3.
Bing Webmaster Tools reports how often Copilot and Bing cite your pages, and its Citation Share gives your slice of all citations for a query, without naming the other sites7. Neither report shows who was cited in your place, so your own sample stays the core measure.
Before you share the results
- Access checked for every assistant
- Same questions, run count and assistants as last round
- Every figure shown with its run count
- Named and cited in separate columns
- Every cited link opened, dead ends marked
- Never-named questions listed with who was cited instead
The tool at the top of this page does this legwork for you, from the crawler check to the ranked list of cited sites.
How it works
Start with the questions your buyers ask and your website. River reads your robots.txt and snippet settings, then asks each question several times in every assistant it can reach without your login. Each answer becomes a row in a spreadsheet: whether you were named, and every source it cited. You get the spreadsheet and a findings page in your workspace.
Every cited site is ranked and marked with the route onto it, or flagged as a dead end. River never invents an answer, a citation or a share, and it tells you which assistants it couldn't reach.
Add your questions
Paste the questions buyers ask before they know you, plus your website and the rivals you lose to.
River asks the assistants
It checks your crawler settings first, then asks each question several times and logs every answer.
Get your report
A spreadsheet of every run and a findings page: named versus cited, top sources and routes.
Act, then rerun
Ask River to draft outreach or write corrected crawler settings, then rerun the same set next quarter.
What you get
- Your robots.txt and snippet settings checked: which AI assistants they shut out, and what that costs
- A spreadsheet in your workspace with one row per answer: question, assistant, run, named, cited links
- How often answers name you, reported apart from how often they cite your pages, with run counts
- Every cited site ranked by runs, with the exact pages behind it and the route onto it
- The questions where you're never named, each listed with the sites the assistants cited instead
- The method written down, with questions, assistants, run count and dates, so you can rerun it exactly
Common questions
What is AI search visibility?
AI search visibility is how often AI assistants such as ChatGPT, Google's AI Overviews and Perplexity name your brand or cite your pages when people ask about your category. Because answers vary between runs, it is best measured as a share of repeated runs of the same questions, with mentions and citations counted separately.
How do I track brand mentions in AI search with River?
Paste your buyers' questions, your website and the competitors you lose to into the box at the top of this page. River checks your crawler settings, asks each question several times in the assistants it can reach, and writes a spreadsheet and a findings page into your free River workspace.
How many times should I run each question?
Five runs per question in each assistant is a practical start for a set of 20 to 30 questions. There's no settled number: SparkToro's study lists it as an open question. Whatever you choose, keep it the same for every question and show it beside each figure. Totals across the set are steadier than any one question's share.
Which AI assistants does River check?
River asks the assistants it can reach on the open web without your login, and the report names every assistant it covered and any it couldn't reach. You can also paste in answers from assistants you use yourself; River keeps those in their own rows, so they're never mixed with its runs.
Will blocking AI training crawlers hide me from AI answers?
Not from ChatGPT search or Google's AI Overviews and AI Mode, as long as only the training crawlers are blocked. OpenAI treats GPTBot (training) and OAI-SearchBot (search) as independent settings, and Google says Google-Extended doesn't affect inclusion in Search. Problems start when one rule blocks every bot or names the search crawler by mistake.
Do I need an llms.txt file or special markup to get cited?
No, not for Google. Its guidance for site owners says no new machine-readable files, AI text files or special schema markup are needed to appear in AI Overviews or AI Mode. The work that helps is ordinary: pages crawlers can reach, plain answers to real questions, and a place on the sites the assistants already cite.
Is the AI search visibility report free?
Yes, it is free to start. Your report and the spreadsheet of runs open in a River workspace, where you can edit them, share them or ask River to change them. Paid plans add more AI use and access to the most advanced models.
Sources
OpenAI Developers. “Overview of OpenAI Crawlers.” developers.openai.com (opens in a new tab)
Google for Developers. “AI Features and Your Website.” developers.google.com (opens in a new tab)
Google. “New opportunities, control and insights for website owners.” blog.google (opens in a new tab)
SparkToro. “NEW Research: AIs are highly inconsistent when recommending brands or products; marketers should take care when tracking AI visibility.” sparktoro.com (opens in a new tab)
Columbia Journalism Review. “AI Search Has a Citation Problem.” cjr.org (opens in a new tab)
Pew Research Center. “Google users are less likely to click on links when an AI summary appears in the results.” pewresearch.org (opens in a new tab)
Microsoft Bing Blog. “New AI Visibility Insights in Bing Webmaster Tools: Intents, Topics, Citation Share, Compare.” blogs.bing.com (opens in a new tab)
Built by
Chandler Supple
Co-founder & CEO, River
Chandler has spent the last six years working in AI, building tools for millions of users and Fortune 100 companies. In 2025, he co-founded River to build a new kind of operating system for AI agents.