How to Get Your Brand Cited by AI (What 1,100 AI Answers Show)

Ask ChatGPT, Perplexity, or Google's AI Overviews to name the best AI meeting notetaker and you'll get a confident, tidy shortlist. We wanted to know how much that shortlist actually favors some brands over others. So over a few days in July 2026 we ran the five biggest AI note-taking tools through all three engines and counted how often each brand showed up.

The spread was wider than we expected. Otter landed in 68.5% of answers. Fathom landed in 28.3%. Same category, same questions, same week. One tool gets named in two-thirds of answers, another in about one in four. That gap is not a verdict on the products. It's a map of who the web talks about, and where the room to grow is.

How we measured this

We scored one thing: does the brand get named in the answer? We set up five projects, one per tool, for Otter, Fireflies, Fathom, Granola, and Jamie. Each ran the same set of buying-intent questions ("best AI meeting notetaker," "Otter vs Fireflies," "AI notetaker for sales calls," and so on) against ChatGPT (GPT-5.3), Perplexity's Sonar, and Google AI Overviews.

Every project scanned three to four times, so these figures come from roughly 1,100 answers, not one lucky run. A mention is simple: the brand name appears in the response text. Citations are a separate signal, and that's where the story gets interesting.

Does publishing more content explain the gap?

Not really. The gap tracks something most brands don't watch: how many different websites cite each tool when AI answers a question. We call it the off-site citation footprint, the count of distinct third-party domains an engine pulls from when your name comes up.

Fathom has the thinnest footprint of the five, around 206 domains, which lines up with its lower mention rate. Otter and Fireflies get cited across 283 domains each and sit at the top. Read the other way, that thin footprint is the clearest lever a brand like Fathom has: the fix is off-site coverage it doesn't have yet, and that's buildable.

It isn't a perfectly straight line. Granola actually has the widest footprint, 288 domains, yet lands mid-pack at 58%. So raw domain count isn't the whole answer. Which sites cite you matters more than how many do. That sets up the real question.

So which sites does AI actually trust?

One site shows up more than any other, for every tool we tested: Reddit.

After we stripped out each tool's own website, reddit.com was the single biggest external citation source in the data, with 334 citations. That's more than Zapier (120) and G2 (92) put together. It wasn't one viral thread carrying the total, either. Reddit was the #1 external source for all five tools individually, from 56 citations for Fathom up to 78 for Granola.

This lines up with what broader studies keep finding. Semrush analyzed more than 100 million AI citations across 230,000 prompts in late 2025 and found Reddit ranking as the most-cited domain on Perplexity and near the top on ChatGPT and Google. Our data says the same pattern holds all the way down inside a single product category.

One honest caveat before you tattoo this on a slide: that same study found Reddit's citation share swings hard week to week, dropping from roughly 60% to 10% on ChatGPT inside a month. Our numbers are a July 2026 snapshot, not a permanent law. Treat the direction as durable and the exact percentages as a moment in time.

Why does one forum beat hundreds of blogs?

Because Reddit is the most citation-dense source we measured. It earns far more citations per domain than anything else.

Blogs produced the most citations overall, about 1,450, but they needed 331 separate domains to do it. Reddit produced 344 citations from six. On a per-domain basis Reddit earns roughly 57 citations each, against 4 for the average blog and 6 for the average listicle. One well-ranked Reddit thread can outweigh a stack of guest posts you spent a quarter placing.

Do the three engines agree?

No, and the difference is worth planning around. Google's AI Overviews named these brands most often, in 59.3% of answers, then ChatGPT at 54.7%, then Perplexity at 50.8%.

Same brands, same questions, three different verdicts. If you only check one engine, you're seeing about a third of your real visibility. How you show up in Google's AI Overviews depends on Google finding supporting pages as it builds the answer, a different sourcing path than a chat model reaching for its most-cited domains. Optimizing for one doesn't guarantee the others.

The part we almost got wrong

Here's a mistake worth sharing, because it changes how you should read any citation report. When we first ranked citation sources per tool, the top source for each one came back as the tool's own website. Fathom's most-cited domain was fathom.ai. Otter's was otter.ai. Accurate, and useless as insight.

Your own site being cited is the price of entry, not an advantage. Once we filtered out every brand's own domain (and its competitors'), the real external winner surfaced, and it was Reddit every time. If your visibility dashboard tells you your homepage is your best-performing citation source, it's flattering you. The sources that swing an answer are the ones you don't own. Earning those is a discipline of its own, and we wrote a full playbook on getting cited off-site by AI for the tactical version.

What to do if you're the one being ranked

The playbook mostly writes itself from the data. Three moves, in rough priority order:

• Get into the roundups the models already pull from. The listicles and review pages that rank for "best AI notetaker" feed retrieval directly. Being absent from them is a citation you're handing a competitor.
• Build a genuine Reddit presence. Not spam. Real answers in the subreddits where your buyers already compare tools, so the thread that gets cited actually mentions you.
• Measure across all three engines, repeatedly. A single scan is a snapshot, and we just showed the engines disagree. Tracking the gap over time, against named competitors, is the only way to know if any of it is working.

That last point is its own skill. Here's how to run competitor analysis for AI search without drowning in dashboards.

The takeaway

On-site fixes are table stakes. Every tool in this study has a clean website, valid schema, fast pages. The 40-point gap between them lives almost entirely off their own domains, in whether the wider web (and Reddit in particular) talks about them. Most brands still treat that part as unmeasurable.

It isn't. You can see your own version of these numbers in a few minutes. Run a free AI visibility scan on your domain and watch which sources AI reaches for when it decides whether to recommend you.