Claude now takes notes on video calls. Searches for in-person note-takers have tripled.

Published
Claude takes notes. What’s left to build? An orange AI figure joins a meeting.

Granola’s estimated visits rose 55% from May to September and searches for in-person AI note-taking more than tripled. Where a new meeting tool could fit.

Claude can now join a meeting, write the notes and start working on the next document before the call ends. For now it joins video calls only, starting with Google Meet, so a conversation in a room is outside it. If you were planning to build an AI meeting app, that is a reason to revisit the pitch.

We pulled the company and keyword histories stored in Night Watcher to see where to look next. Two findings stood out: Granola’s estimated monthly visits grew 55% between May and September, and searches for “ai note taker for in person meetings” rose from 390 to 1,300 year over year.

That suggests a direction worth investigating: the way a product captures a conversation may offer a more specific starting point than another general-purpose summary app.

These observations end in September, before Claude Meetings launched on November 12. They show the market heading into the launch; they cannot tell us how customers have responded to it.

Granola grew faster, but the bigger domains stayed bigger

Granola’s domain went from an estimated 1.88 million monthly visits in May to 2.91 million in September. That was the strongest growth in our four-domain comparison.

Fathom’s fathom.video domain grew about 16%. Otter and Fireflies each fell about 4%, less than the 8% median decline among tracked sites their size. This was an uneven market, rather than a rising tide lifting every meeting tool.

Monthly traffic indexed to May 2026 equals 100: granola.ai ends at 155, fathom.video at 116, otter.ai at 96 and fireflies.ai at 96.
Source: Night Watcher. Research retrieved November 16, 2026. Domain visits across traffic channels, not revenue or customer counts. The index compares growth rates; the table below shows size.
Tracked domainMay 2026 visitsSeptember 2026 visitsChange
granola.ai1,883,3612,914,683+54.8%
fathom.video4,838,8035,609,517+15.9%
otter.ai6,972,4176,716,531−3.7%
fireflies.ai4,457,8524,284,664−3.9%

Granola grew fastest from a smaller base. Otter still had more than twice its estimated visits in September. Growth and size answer different questions, and neither tells us which business is most profitable.

We also found a separate fathom.ai history: roughly 809,000 to 919,000 visits over the same period, up 13.6%. We have kept it outside the chart and have not added the two Fathom domains together. Separate domain estimates are not a deduplicated audience.

For a founder, Granola is worth studying. The traffic history alone cannot establish why it grew, whether visitors converted or which feature attracted them.

The sharper search growth was in how people capture conversations

The broad phrase “ai note taker” recovered from an estimated 18,100 monthly searches in June to 27,100 in September. But September was still 33% below the same month a year earlier. Calling that an uninterrupted boom would miss the longer view.

Two smaller phrases moved differently:

  • “ai note taker device” rose from 880 in June to 1,600 in September, about 82%. It was up 233% year over year.
  • “ai note taker for in person meetings” rose from 720 to 1,300 over those months, about 81%. It was also up 233% year over year.

Meanwhile, “ai transcription” fell from 5,400 in June to 4,400 in September and was down about 19% year over year.

Four keyword histories from January to September 2026: broad AI note-taking demand recovered, device and in-person queries increased from smaller bases, and AI transcription declined. Each panel uses a separate vertical scale.
Source: Night Watcher. Values are rounded monthly search estimates. Each panel has its own vertical scale, and related queries must not be added together.

The smaller phrases are interesting because they describe a constraint: someone wants to capture a conversation in a room, or is looking for a physical device to do it. They give us a more focused customer problem to investigate than “people need meeting notes.”

They do not tell us that a new device would sell. Searchers could be comparing established products, looking for a free solution or troubleshooting one they already own. A growing query is a lead for customer research, not a forecast of paying customers.

Claude’s launch makes the capture question more useful

Anthropic’s launch announcement describes a visible meeting participant that takes notes and helps create work during a call. The beta is rolling out to Team and Enterprise plans, starting with Google Meet.

That gives founders a concrete comparison to make: what does the customer need outside that setting? A room-based interview, a phone conversation and a laptop video call can have different capture requirements. Those differences are worth testing with the people having the conversations.

Existing specialists already serve parts of this territory. Fathom’s site lists in-person meetings, flexible capture and an iOS app. Granola lists mobile and watch capture alongside its meeting tools. Our finding is a reason to examine this market more closely, not a claim that these needs are unserved.

Nor must a specialist compete with Claude at every step. Fathom’s Claude integration exposes meeting transcripts, summaries and action items through MCP. Granola also makes meeting context available to AI applications. A product can capture useful context and feed the assistant a customer already uses.

The first product test we would run

We would start with one recurring in-person conversation and one job that follows it. For example, a product researcher might need interview evidence attached to the right research question, with the original passage available for review.

That is a proposed test, not a validated opportunity. Before building, observe several real sessions with permission and record where the current process breaks: capturing the audio, identifying the speaker, finding a passage or filing the result.

Then ask participants to complete the same job using their existing tools. Measure time spent correcting the output and moving it into the destination system. If an incumbent already handles it well, move on. If a repeated problem remains, test whether solving it earns a paid pilot.

The search history helps choose whom to interview. The company history helps choose which competitors to study. Customer behavior has to answer whether there is a business.

Our public profiles of Granola, Otter and Fireflies collect what each has said about valuation, funding and revenue, with sources.

How we checked the numbers

We retrieved stored records from Night Watcher on November 16, 2026. The traffic comparison uses all five available consecutive months, May through September, for four selected meeting-tool domains. The data is modeled domain traffic, not first-party analytics or US-only visits.

For search demand, we inspected 60 canonical keyword records from a targeted meeting-note and transcription search, then compared four relevant phrases using their January–September histories, June-to-September change and September year-over-year change. This is a selected comparison, not an exhaustive survey of every meeting-related keyword. We excluded duplicate spellings and did not sum overlapping queries.

The company snapshots were refreshed in October; the underlying months still end in September. Neither dataset measures the effect of Claude’s November launch. We have not tested transcription accuracy, product reliability or willingness to pay.

Jordan Cole

Creator of NightWatcher AI. Specializes in data-driven insights for AI product development, market validation, and competitive analysis.