MemetikEdition 2026-09

Lists / Startup stack

Best AI meeting notetakers for founders (2026): What ChatGPT, Claude & Gemini Recommend

Fathom appears in 48 of 50 AI answers, but Granola is named first most often. A ranked list of 19 AI meeting notetakers, with the per-model split for founders.

Fathom is the AI meeting notetaker that AI models name most often for founders. It appears in 48 of 50 recorded answers and comes first in 8 of them. Granola is the one they open with. It comes first in 19 of 50 answers, and the founder answers keep giving one reason: no bot joins the call. Fireflies is the usual lead pick for sales calls that need CRM sync.

Most tools in this category are AI meeting notetakers that join your calls, transcribe them and generate summaries.

The counts measure which products the models name, not how well the products work.

TL;DR

1. Fathom

Pick Fathom if you want a free notetaker that every model puts on its list, especially for Zoom-heavy sales and customer calls.

Measured: named 48 of 50 (Fathom 96%), first in 8 of 50 (Fathom 16%), average position 3.04.

Fathom is the constant in this panel. Eight of the ten models name it in all five answers, and ChatGPT and Sonar Reasoning Pro name it in four. The models rarely open with it, though. Their answers cast Fathom as the free default placed beside a first choice. Perplexity put it plainly: “If your top priority is a free option, Fathom is the strongest budget-friendly pick.”

Nadia Okafor’s Medium test says Fathom is free and good enough for solo Zoom use.

Read AI’s guide says Fathom is built primarily for sales professionals.

Pros

Cons

Pricing: Free plan available. Paid team plans start at $19 per user per month, per Read AI’s guide.

Best for: founders on Zoom who want a no-cost default, and sales calls that need CRM notes.

2. Fireflies

Choose Fireflies if your notes have to land in a CRM after sales calls, because that is the question the models tie it to.

Measured: named 47 of 50 (Fireflies 94%), first in 4 of 50 (Fireflies 8%), average position 4.09.

Fireflies sits on nearly every list and at the top of very few. The founder questions treat it as a second name. The sales question changes that. Claude Opus 5, Claude Fable 5, Perplexity and Sonar Reasoning Pro all open their sales-and-CRM answer with Fireflies. Claude Opus 5 called it “the most consistent recommendation for CRM sync specifically”. A founder buying for personal notes will find the models pointing elsewhere. A founder running the sales motion will find Fireflies in almost every answer and leading the CRM ones.

Okafor calls Fireflies the strongest bot based pick if your team needs CRM sync. She says it plugs into Salesforce, HubSpot, Pipedrive and several others.

Pros

Cons

Pricing: Free plan with limited storage. Pro plan starts at $10 per user per month, per Read AI’s guide.

Best for: founder-led sales teams that want call notes pushed into the CRM.

3. Granola

Pick Granola if you are the founder taking investor, hiring and customer calls yourself and want notes without a bot in the room.

Measured: named 39 of 50 (Granola 78%), first in 19 of 50 (Granola 38%), average position 2.79.

Granola ranks third on presence and first on commitment. The founder wording drives it. GPT-5.6 Sol, GPT-5.6 Luna, Claude Opus 5 and Claude Fable 5 each lead a founder answer with Granola, and the reason they give is the missing bot. Some of that agreement is literal. Claude Fable 5 wrote “Granola is the 2026 darling of the AI note-taking world” across its founder answers, and Claude’s first founder answer carries the same phrase. Each of those answers cites simular.ai, the most-cited host in the category.

Okafor reports that Granola doesn’t save audio, only transcripts, so a mistranscription can’t be checked by listening back. She finds its integrations weaker than Fireflies.

Read AI’s guide notes there is no Android app yet.

Pros

Cons

Pricing: Business plan at $14 per user per month, per Okafor. Read AI’s guide lists a free plan capped at 25 meetings lifetime and an Individual plan at $18 per user per month.

Best for: founders running investor, candidate and customer conversations who do not want a visible recorder.

4. Zoom

Choose Zoom’s built-in assistant if your calls already run on Zoom and you would rather not add another subscription. Read its rank with care.

Measured: named 31 of 50 (Zoom 62%), first in 3 of 50 (Zoom 6%), average position 4.26.

Zoom collects two kinds of mention. The panel matches both “Zoom AI Companion” and plain “Zoom”, so an answer saying a notetaker works on Zoom counts the same as one recommending Zoom’s own assistant. Both kinds occur. GPT-5.6 Luna lists Zoom AI Companion as the value choice for teams already on a major meeting platform. ChatGPT’s first founder answer uses the word only to list the platforms Fathom supports. Gemini and Gemini 3.5 Flash name Zoom in every answer, a pattern that fits answers listing supported platforms, although the count alone cannot confirm it. No captured guide reviews Zoom’s assistant.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: teams already on Zoom that want notes without a separate tool.

5. Microsoft Copilot

Pick Microsoft Copilot if your company already runs on Microsoft Teams, and discount its count before comparing it with the specialists.

Measured: named 30 of 50 (Microsoft Copilot 60%), first in 1 of 50 (Microsoft Copilot 2%), average position 5.43.

Microsoft Copilot carries the same caution as Zoom, only stronger. The panel counts both “Copilot” and “Teams” as Microsoft Copilot, so any answer saying a product works with Microsoft Teams registers a Copilot mention. ChatGPT’s first founder answer praises Fathom for working across Zoom, Google Meet and Microsoft Teams. One compatibility line adds to three platform rows at once. The spread across models is uneven. Gemini 3.5 Flash names it in every answer, and Claude Fable 5 names it in 1 of 5. With one first placement and a 58-point gap between named and first, the models list it far more often than they choose it.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: Microsoft 365 companies that want notes inside Teams.

6. Otter

Choose Otter if you record in-person meetings on your phone and want a free monthly allowance, after weighing the lawsuit Okafor flags.

Measured: named 27 of 50 (Otter 54%), first in 2 of 50 (Otter 4%), average position 5.04.

Otter is the OpenAI models’ product. The three GPT models name it in most of their answers, and ChatGPT’s comparison answer ranks Otter.ai best overall for most people. Outside that family it fades. Claude Fable 5 and Gemini 3.5 Flash name it once each. A founder who asks an OpenAI model will meet Otter often. A founder who asks any other model may not see it at all.

Okafor writes that Otter’s iOS app remains the best for in person meetings or recording yourself talking. She reports a class action filed in California in August 2025, Brewer v. Otter.ai, alleging that Otter recorded conversations and used them to train its models without proper consent from all participants. She notes the case is still in early stages.

Pros

Cons

Pricing: Free plan with 300 minutes a month. Paid plans start at $8.33 per user per month, billed annually, per Read AI’s guide.

Best for: founders who record in-person conversations on a phone.

7. Notion

Consider Notion only if your company already works in Notion and wants notes in the same place, because the models mostly name it as a destination.

Measured: named 25 of 50 (Notion 50%), first in 0 of 50 (Notion 0%), average position 4.44.

Notion never opens an answer. The panel matches “Notion AI Meeting Notes” and plain “Notion”, and plain Notion turns up constantly as an integration target. GPT-5.6 Sol’s founder answer, for example, lists Notion among the tools Granola integrates with. That mention counts exactly like a recommendation. The split by model is wide. Neither Perplexity model names Notion at all, while GPT-5.6 Luna, Gemini and Gemini 3.5 Flash name it in 4 of 5 answers each. A model that lists integrations will name Notion whether or not it rates Notion’s own notes. No captured guide reviews Notion’s meeting notes.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: Notion-first teams that want meeting notes in the same workspace.

8. Google Meet

Use Google Meet’s own notes if every call you take already runs on Meet, because its count rests mostly on one model’s answers.

Measured: named 17 of 50 (Google Meet 34%), first in 0 of 50 (Google Meet 0%), average position 5.35.

Google Meet has the most lopsided split of the platform entries. Gemini 3.5 Flash names it in all 5 answers, while GPT-5.6 Sol never does. Like Zoom, it is matched by platform name as well as by “Gemini in Meet”, so an answer listing supported platforms counts. For a founder who lives in Meet, the built-in option and the bot-free tools are the ones to compare first, for the reason Okafor gives.

Okafor reports that as of March 2026, Google flags third party note taker bots as potential risk by default in Google Meet. In her account, the host has to manually let them in, every time.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: Google Workspace teams that run every call in Meet.

9. Fellow

Pick Fellow if you are buying for a team with compliance requirements rather than for yourself.

Measured: named 16 of 50 (Fellow 32%), first in 0 of 50 (Fellow 0%), average position 6.69.

Fellow is an Anthropic-family name in this panel. Claude names it in 4 of 5 answers and the family figure is Fellow 60%, while none of the three OpenAI models names it once. It never opens an answer and sits low when it appears. Okafor’s test explains who it serves.

Okafor calls Fellow the choice for organisations with compliance requirements. She calls it the safest pick for a team that all needs to be on the same tool with shared admin oversight. Of the team-focused tools, she says that if you’re choosing for yourself, you probably don’t need any of them.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: teams that need admin controls and compliance coverage across every seat.

10. Avoma

Choose Avoma if sales calls feeding a CRM are the main job, because the sales-and-CRM question is where the models lead with it.

Measured: named 15 of 50 (Avoma 30%), first in 3 of 50 (Avoma 6%), average position 5.6.

Avoma’s first placements all come from one question. ChatGPT, GPT-5.6 Sol and GPT-5.6 Luna each open their sales-and-CRM answer with Avoma, and each gives CRM automation as the reason. Gemini 3.5 Flash never names it. The models treat Avoma as a sales-operations tool, not a founder’s personal notetaker, and the counts only make sense read that way. The one captured review is a sales-notes test from Arrows.

Arrows’ reviewer said Avoma’s action items were not super specific. The same reviewer said the notes did not capture the action items expected from the prospect.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: sales teams that want CRM fields filled from calls.

11. Jamie

Pick Jamie if privacy is the deciding factor and you want bot-free notes that run locally.

Measured: named 14 of 50 (Jamie 28%), first in 4 of 50 (Jamie 8%), average position 3.57.

Jamie splits the panel by model family. Perplexity and Sonar Reasoning Pro name it in 4 of 5 answers each, and both make it their lead pick in at least one founder answer. The three OpenAI models and Gemini never name it. When it does appear, it sits high. A source pattern sits behind the split. The Perplexity-family answers that lead with Jamie cite Jamie’s own blog, and meetjamie.ai is cited 37 times across the recorded answers in this category.

Okafor describes Jamie as bot free, running locally, working offline and supporting a hundred plus languages. She rates its privacy posture stronger than Granola’s.

Pros

Cons

Pricing: The free plan gives ten summaries a month, per Okafor. Paid pricing is not recorded in the captured pages.

Best for: privacy-first founders who want bot-free capture.

12. Gong

Choose Gong if you are equipping an enterprise sales team, not taking your own founder notes.

Measured: named 13 of 50 (Gong 26%), first in 1 of 50 (Gong 2%), average position 4.62.

Gong appears across the whole panel, but thinly. Every model names it at least once and none more than 3 of 5 times. The recorded answers agree on its role. Claude opens its sales-and-CRM answer with Gong as the choice for enterprise sales teams that need deep conversation intelligence. Claude Fable 5 places it as the option for enterprise teams and managers. Perplexity positions it toward conversation intelligence and team analytics rather than lightweight note-taking. None of the captured guides covers Gong, so there is no sourced pricing or feature record for it. A founder choosing a personal notetaker will find the models pointing elsewhere.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: enterprise sales teams buying conversation intelligence.

13. Laxis

Look at Laxis only if you want a bot-free sales notetaker with HubSpot or Salesforce sync, and read its visibility with its own marketing in mind.

Measured: named 11 of 50 (Laxis 22%), first in 2 of 50 (Laxis 4%), average position 5.36.

Laxis has an unusual source profile. Its own site, laxis.com, is cited 54 times in the recorded answers, more than any other vendor-owned host. Claude Opus 5 names it in 3 of 5 answers and also warns about “the Laxis post that crowns Laxis #1”. Sonar Reasoning Pro offers Laxis as a lower-priced, bot-free alternative with native HubSpot and Salesforce sync for sales calls. The OpenAI models and Gemini never name it, and none of the captured guides reviews it. A buyer should read its presence as partly the reach of its own content. That is a visibility fact, not a verdict on the product.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: sales-led founders who want bot-free capture with CRM sync.

14. tl;dv

Pick tl;dv if you want a generous free tier and plan to reuse call clips to train people.

Measured: named 9 of 50 (tl;dv 18%), first in 1 of 50 (tl;dv 2%), average position 5.78.

tl;dv gets most of its support from Claude Opus 5, which names it in 3 of 5 answers. ChatGPT, Claude, Claude Fable 5, Gemini and Gemini 3.5 Flash never name it. Claude Opus 5 also points out that tl;dv’s own post names tl;dv best overall, which is worth knowing when that post turns up in search.

Okafor finds tl;dv closer to a meeting clip and library tool than a pure note taker, which is useful if you’re training other people on calls. She reports a recently added desktop app that captures audio without a bot.

Pros

Cons

Pricing: Free plan with unlimited meetings. Paid plans start around $9.99 per user per month, per Read AI’s guide.

Best for: teams that clip and share recorded calls.

15. Read AI

Consider Read AI if you want meetings, email and messages searchable in one place, and weigh that its strongest advocate is its own guide.

Measured: named 8 of 50 (Read AI 16%), first in 0 of 50 (Read AI 0%), average position 7.88.

The models are cool on Read AI. None of the three Anthropic models names it, it never opens an answer, and it sits deep in the answers that include it. The read.ai domain is cited 34 times in the recorded answers, so its own pages reach the models even when the product stays low. ChatGPT’s comparison answer is the exception, naming it the pick for meeting analytics and knowledge search.

Read AI’s own buying guide ranks Read AI at number 1 for startups.

Okafor found its extra features more distracting than useful.

Pros

Cons

Pricing: Free plan with 5 meetings a month. Pro plan runs $15 per user per month billed annually, per Read AI’s guide.

Best for: teams that want meeting notes joined to email and chat search.

16. Shadow

Try Shadow if you run a Mac-based founding team and want capture that starts without a bot.

Measured: named 7 of 50 (Shadow 14%), first in 2 of 50 (Shadow 4%), average position 4.29.

Shadow is a Perplexity-side product. Perplexity and Sonar Reasoning Pro each name it in 2 of 5 answers, and both open their first founder answer with it. Perplexity called it the strongest pick for a Mac-based founding team that wants bot-free, auto-detected capture with screen-aware notes. No OpenAI model names it, and neither do Claude Opus 5 or Claude Fable 5. Its own site, shadow.do, is cited 28 times in the recorded answers, and both Perplexity founder answers that lead with Shadow cite a shadow.do blog post. None of the captured guides reviews Shadow, so the record holds nothing on pricing or features beyond what the answers say.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: Mac-based founding teams.

17. Grain

Consider Grain if you run customer-facing sales calls and want short notes, knowing that one model carries half its count.

Measured: named 4 of 50 (Grain 8%), first in 0 of 50 (Grain 0%), average position 6.25.

Grain is close to a ChatGPT-only name. ChatGPT names it in 2 of 5 answers, and its comparison answer calls it the pick for sales and customer-facing teams. Claude Opus 5 and Claude name it once each, and no other model does. The one captured review is the Arrows sales-notes test.

Arrows’ reviewer said Arrows uses Grain itself. The reviewer called Grain’s notes really nice and concise, with good action items. The reviewer judged them not quite as thorough as Fathom’s notes.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: sales and customer-success teams that want short call notes.

18. Attention

Attention suits only a buyer who already has it on a shortlist from elsewhere, because this panel adds almost no evidence either way.

Measured: named 1 of 50 (Attention 2%), first in 0 of 50 (Attention 0%), average position 9.

One answer names Attention, and it comes from Claude Opus 5, the only model to name Attention or Circleback. That answer places it ninth. No other model mentions it, none of the captured guides reviews it, and the record holds no pricing or features. There is a matching caveat too. Attention is an ordinary English word, and the panel counts names by string match. A single hit cannot be confirmed as a product mention without reading the answer in the full category record. The fair reading is that the models do not currently treat Attention as a founder notetaker, whatever its merits for other jobs.

Pros

Cons

Pricing: No public pricing is recorded in the captured pages.

Best for: the record holds no use case to state.

19. Circleback

Circleback suits a buyer starting from scratch who will test tools personally, because the testers rate it and the models barely name it.

Measured: named 1 of 50 (Circleback 2%), first in 0 of 50 (Circleback 0%), average position 12.

Circleback shows the widest split between the captured guides and the models. The models almost ignore it. Claude Opus 5 names it once, twelfth in that answer, and no other model mentions it. For a buyer, the gap is the finding. People who tested Circleback rate it, while the answers a founder reads in an AI tool rarely surface it.

Arrows’ reviewer named Fathom, Granola and Circleback as the top three favourites for the quality of sales notes.

Okafor calls Circleback newer, with reasonable integrations and competitive pricing. She would consider it if starting from scratch, and sees no reason to switch from a tool you already like.

Pros

Cons

Pricing: Described as competitive by Okafor, with no figure given.

Best for: buyers choosing a first notetaker who will run their own trial.

How the tools compare

Fathom leads on presence, Fireflies sits one answer behind, and Granola is ahead of both on first placements.

Vendor Named Share First First share Avg position Pricing model in captured guides
Fathom 48/50 Fathom 96% 8/50 Fathom 16% 3.04 Free plan, paid team seats
Fireflies 47/50 Fireflies 94% 4/50 Fireflies 8% 4.09 Free plan, paid per user
Granola 39/50 Granola 78% 19/50 Granola 38% 2.79 Lifetime-capped free plan, paid per user
Zoom 31/50 Zoom 62% 3/50 Zoom 6% 4.26 Not recorded
Microsoft Copilot 30/50 Microsoft Copilot 60% 1/50 Microsoft Copilot 2% 5.43 Not recorded
Otter 27/50 Otter 54% 2/50 Otter 4% 5.04 Free minutes, paid per user
Notion 25/50 Notion 50% 0/50 Notion 0% 4.44 Not recorded
Google Meet 17/50 Google Meet 34% 0/50 Google Meet 0% 5.35 Not recorded
Fellow 16/50 Fellow 32% 0/50 Fellow 0% 6.69 Not recorded
Avoma 15/50 Avoma 30% 3/50 Avoma 6% 5.6 Not recorded
Jamie 14/50 Jamie 28% 4/50 Jamie 8% 3.57 Free plan, paid not recorded
Gong 13/50 Gong 26% 1/50 Gong 2% 4.62 Not recorded
Laxis 11/50 Laxis 22% 2/50 Laxis 4% 5.36 Not recorded
tl;dv 9/50 tl;dv 18% 1/50 tl;dv 2% 5.78 Free plan, paid per user
Read AI 8/50 Read AI 16% 0/50 Read AI 0% 7.88 Free plan, paid per user
Shadow 7/50 Shadow 14% 2/50 Shadow 4% 4.29 Not recorded
Grain 4/50 Grain 8% 0/50 Grain 0% 6.25 Not recorded
Attention 1/50 Attention 2% 0/50 Attention 0% 9 Not recorded
Circleback 1/50 Circleback 2% 0/50 Circleback 0% 12 Not recorded

Share is the share of the 50 recorded answers that named the product. First share is the share where it appeared before any other tracked product. The pricing column condenses the plans quoted in each entry.

Below the top three, the table sorts into three groups. Specialist notetakers that the models sometimes open with: Jamie, Avoma, Shadow and Laxis. Platforms and workspaces counted by name: Zoom, Microsoft Copilot, Notion and Google Meet, which appear often and open almost nothing. Then a tail named by one or two model families, from Fellow to Circleback. Wispr was tracked and never named.

Why do Zoom, Microsoft Copilot and Notion rank above the specialist notetakers?

They rank high because the panel counts names, and those names appear in answers as platforms and integrations as often as recommendations. The table cannot show that on its own.

The panel matches each product by a short list of names. Zoom matches “Zoom” as well as “Zoom AI Companion”. Microsoft Copilot matches “Copilot” and “Teams”. Google Meet matches “Google Meet” and “Gemini in Meet”. Notion matches “Notion”. A notetaker answer uses these words for ordinary reasons: to list the meeting platforms a tool supports, or the workspaces it exports to.

Two recorded answers show the mechanism. ChatGPT’s first founder answer recommends Fathom and says it works across Zoom, Google Meet and Microsoft Teams. That single compatibility line counts toward three platform rows. GPT-5.6 Sol’s founder answer lists Notion among the tools Granola integrates with, which counts toward Notion. Neither sentence recommends the platform.

Genuine recommendations exist as well. GPT-5.6 Luna names Zoom AI Companion and Google Meet’s note-taking as the value choice for teams already on a major meeting platform. The count cannot separate the two uses.

The first column gives it away. Notion and Google Meet open none of the 50 answers, and Zoom and Microsoft Copilot open very few. A product the models chose would open answers. For a founder, rows 4, 5, 7 and 8 mean “check the assistant in the tools you already pay for”. They do not add four more specialist notetakers to evaluate. The answers are readable in full in the category record.

Where do the models disagree?

They agree on the leader and disagree on almost everything under it. All ten models name Fathom at least as often as any other product, and every model family has Fathom as its leader.

The question moves the pick more than the model does. ChatGPT opened its answer to the first question with Fathom, the second and fourth with Granola, the third with Otter and the fifth with Avoma. One model, five questions, and the lead changes with the wording. The founder questions pull towards Granola. The sales-and-CRM question pulls towards Fireflies, Avoma and Gong.

Anthropic’s models favour Granola and the team tools. Within the Anthropic family the figures are Granola 93.3%, Fireflies 100% and Fellow 60%, and Fellow is a product no OpenAI model names. Claude Opus 5 is the only model to name Attention or Circleback. It also warned about its own sources: “nearly every “best AI notetaker” listicle is published by a vendor in the space, so treat any single ranking with skepticism”.

OpenAI’s models favour Otter. Otter 73.3% across the three GPT models, against a single mention each from Claude Fable 5 and Gemini 3.5 Flash. None of the three GPT models names Jamie or Fellow.

Google’s models name the platforms. Zoom 100% and Granola 100% across Gemini and Gemini 3.5 Flash, with Microsoft Copilot 90%. Gemini 3.5 Flash names Google Meet in every answer.

Perplexity’s models favour Jamie and Shadow. Jamie 80% within that family, and Granola only 50%. Neither Perplexity model names Notion. Their founder answers cite vendor blogs, including meetjamie.ai and shadow.do. The most-cited vendor-owned hosts in this category are laxis.com (54), meetjamie.ai (37), read.ai (34), shadow.do (28) and granola.ai (16). The most-cited host overall is simular.ai (116).

How was the sample built?

The sample is 10 models x 5 fixed prompts = 50 recorded answers, run on 2 September 2026 for edition 2026-09. Each model answered each question once. The five questions, verbatim:

  1. “What is the best AI meeting notetaker for a founder? Name specific products.”
  2. “Which AI meeting notetaker would you recommend to a founder in 2026?”
  3. “Compare the top AI meeting notetaker options right now.”
  4. “I’m a founder and I need an AI meeting notetaker. What should I use and why?”
  5. “Best AI meeting notetaker for sales calls with CRM sync?”

The ten models come from OpenAI, Anthropic, Google and Perplexity. OpenAI supplies GPT-5.6 Sol, ChatGPT (served by GPT-5.6 Terra) and GPT-5.6 Luna, for 15 answers. Anthropic supplies Claude Opus 5, Claude (served by Claude Sonnet 5) and Claude Fable 5, for 15 answers. Google supplies Gemini (served by Gemini 3.6 Flash) and Gemini 3.5 Flash, for 10 answers. Perplexity supplies Perplexity (served by Sonar Pro) and Sonar Reasoning Pro, for 10 answers.

The panel tracked 20 products and 19 were named. A product counts as named when one of its listed names appears in the answer. Named first means it appeared before any other tracked product. The full method is on the method page.

How do these counts compare with the AI meeting notetaker guides?

The pages that rank for this question are buying guides. They rank products for purchase. The counts record the names AI answers produce. The two overlap at the top and part ways lower down.

Nadia Okafor’s Medium test is the independent one. It covers fourteen tools tested across ninety days of real calls. It states that nobody pays her to write about any of this: no sponsors, no affiliates, no free products. It calls Granola the cleanest tool for notes, Fireflies the strongest bot based pick for CRM sync and Fathom free and good enough for solo Zoom use.

Those three are the panel’s top three.

Read AI’s startup guide is a vendor ranking its own product. Titled “8 Budget-Friendly AI Meeting Assistant Tools for Startups”, it lists free-plan limits and prices for each of its eight tools. Its plan limits and prices are useful, and the entries quote them with that attribution.

Arrows’ YouTube test judges note output on a sales call. The reviewer ran 22 AI notetakers on the same call script and compared only the quality of the notes. Arrows sells AI-powered sales rooms, which the video promotes. The video was uploaded on 6 March 2025.

None of the three measures what AI models say. The counts add that layer: which products the models name, which they open with, how the pick shifts by model and by question, and which rows are inflated by platform names. The guides judge products. The panel records the answer a founder gets by asking an AI model instead of reading a guide.

What should a founder do with these counts?

Start with the fork that both the guides and the models use: bot or no bot. Then match the shortlist to the calls you actually take.

Treat any single guide, including a vendor’s own, as one input. The per-model split is a quick check on any name. A product named by every model family is a broad signal. A product named by one family, and cited from its own blog, is a narrow one.

What can these counts not tell you?

They cannot tell you which product is better. Answer share and named first measure presence in recorded answers, not quality, reliability, support, pricing fairness or fit with your stack. Being named is not the same as being recommended, because an answer can list a product only to warn against it.

The sample has fixed limits. There is one response per prompt-and-model pair, so a single answer can move a small vendor’s count. It is one dated snapshot from 2 September 2026. Names are matched as strings, which inflates platform rows and can miscount ordinary words such as Fellow, Shadow, Grain and Attention. API answers can differ from what the same model says in its consumer chat app. The prompts are in English. No vendor paid to appear, be reordered or be removed, and products that were never named are not listed.

Frequently asked questions

Which AI note taker is best for meetings?

No count can settle “best”, but the models’ answer is clear. Fathom is the name they give most often, and Granola is the one they open with most often. For a founder taking sensitive calls personally, the models lean to Granola because no bot joins the call. For sales calls that must reach a CRM, they lean to Fireflies and Avoma. Match the pick to how your meetings run, then trial it.

Are AI notetakers illegal?

The panel does not measure legality and cannot answer the question for your situation. The legal position depends on where you and the other participants are, so check local recording-consent rules before you record anyone.

Okafor reports a class action filed in California in August 2025 alleging that Otter used recorded conversations to train its models without proper consent from all participants. She notes the case is still in early stages.

What is the best note-taking app for in-person meetings?

The panel did not ask about in-person meetings, so its counts do not rank tools for them. The captured guides point to phone apps.

Okafor rates Otter’s iOS app the best for in person meetings or recording yourself talking.

Read AI’s guide says its iPhone and Android apps capture in-person meetings. The same guide notes Granola has no Android app yet.

Is Fathom actually free?

Fathom has a free plan, and it is the product every model in the panel names.

Okafor describes unlimited recordings on Zoom for nothing.

Read AI’s guide lists paid team plans from $19 per user per month.

Check the current plan page before relying on either figure.

Did the AI models test these products?

No. The models answered five buyer questions, often drawing on web pages they cited. A product’s count shows how often those answers named it. It says nothing about how well the product works for you.

Why are Soda, Bluedot and Krisp missing from the ranking?

The panel tracks a fixed list of 20 products, and only those are counted. Tools outside that list are not ranked, even when an answer or a guide mentions them.

Okafor’s test covers all three, and she places Soda in a different category from meeting note takers.

Can a vendor pay to rank higher on this list?

No. Positions are not sold, sponsored or influenced. The order is the measured order of names in the recorded answers, and it changes only when a new edition’s answers change.

How this list is ordered

The order is the measurement, not an assessment of the products. Answer share is the share of recorded answers that named the tool. Named first is the share where it appeared before any other tracked tool. Both are counts from one dated edition and are published in full on the category page.

A tool appears here only if it was named in the edition and its record carries a sourced claim. A product that was never named is not listed, and no position is sold.

Where to check it