I priced Plaud, Granola and Fathom on the same morning. Here’s the first-year number.

The first thing I learned while pricing this out was that fathom.video now 301-redirects to fathom.ai. Small thing. It broke my first fetch, and it’s a fair signal of how fast this category churns.

So I did the boring version of Plaud vs Granola vs Fathom instead. On 20 August 2026 I opened four vendor pages myself (Plaud’s plan pricing page, Plaud’s product listing, Granola’s pricing page, Fathom’s pricing page), copied the figures straight off them, and built the first-year table further down from those figures. Dates are stamped on everything below. If you’re reading this months later, go re-check them.

What Plaud vs Granola vs Fathom costs, as listed on 20 August 2026

These come from each vendor’s own page on that date. Where a page didn’t state something, I’ve left it out rather than filled it in.

Three footnotes on that table, because the team plans muddy it. Fathom also sells a Team plan at $19 a month per user, or $15 annually, with a two-user minimum. Plaud’s Team plan is $28 per user monthly under a launch offer the page says runs to 31 August 2026, going to $35 after that, with annual billing at $20 per user in the first year and $25 later. And Granola’s page showed no annual option at all when I looked, so a year of Business is simply twelve payments of $14.

Something worth flagging about the pages themselves, since it shaped how this got researched. Plaud’s plan pricing page lists the subscriptions but not the devices, so you can’t work out what Plaud costs from the pricing page alone. You have to go to the shop listing. Granola’s page carried no annual price, which may just mean one is on the way. Fathom’s Enterprise tier is contact-sales, which is completely normal and also means the top of its range is unknowable from outside. Three different kinds of gap. Only the first cost me an extra page load, which is why the device prices here come off the shop rather than the pricing page.

Plaud charges you twice, and that changes the math downstream

Granola and Fathom are software. Install, pay monthly, or don’t pay. Plaud is a recorder you buy, plus a subscription that turns what it records into something readable.

Device prices from Plaud’s own product listing on 20 August 2026: Plaud Note at $159.00 USD, Plaud NotePin at $159.00 USD, Plaud Note Pro on sale at $159.99 USD down from a $189.00 list, and Plaud NotePin S on sale at $151.99 USD down from $179.00. Sale prices are sale prices. I wouldn’t plan a budget around them holding.

Which means the real entry cost for Plaud isn’t $17.99. It’s roughly $160 before you have transcribed a single minute, and the subscription starts on top of that. Against Granola Business at $14 a month, you are eleven months into a Granola subscription before you have spent what the Plaud device alone costs.

That’s not a knock on the hardware. A recorder in your shirt pocket does the one thing neither piece of software can do, which is capture a conversation with no laptop anywhere in the room. Corridor conversations. Site walks. A consultation with a doctor. If that is the job you’re hiring for, the device cost is just the cost of entry and this comparison ends right here.

Three products, three different places to put the free ceiling

This was the part I found genuinely interesting, and you only see it by reading the three free tiers side by side rather than one at a time.

Plaud caps minutes. Its free Starter plan gives you 300 transcription minutes a month, which is five hours, and past that the transcription stops until the month rolls over.

Granola caps history. Basic is free and lists AI meeting notes, note templates, shared folders, AI chat across meetings and multi-language support, but meeting history on that tier is limited. You keep the notes from this week. Your archive is the thing you’re buying back.

Fathom caps the intelligence layer. Its free plan lists unlimited recordings and transcriptions, instant AI call summaries, plus clips, playlists and search across calls. What’s held back on free is AI action items and usage of its assistant. So you keep everything you record, and you do less with it.

Those are three different bets about what a person will eventually pay for. Plaud bets you’ll pay for volume. Granola bets you’ll pay to keep your past. Fathom bets you’ll pay for the layer sitting on top of a pile it lets you build for nothing. My guess, and I might be wrong, is that Fathom’s is the strongest of the three, because unlimited free recording gets very hard to walk away from once four months of your calls are sitting inside it. Switching cost you built yourself, for free.

The practical read is that each free tier tells you in advance how it will break. Plaud’s breaks in the fourth or fifth hour of a busy week. Granola’s breaks the first time you go hunting for something you discussed in March. Fathom’s breaks when you get tired of typing your own action items. Pick the one whose breaking point you’d mind least, then treat the paid tier as insurance against that one specific annoyance rather than as a general upgrade.

If you want the wider set of no-bot options rather than just these three, we went through them in the no-bot notetaker roundup.

Where the audio actually comes from

Pricing pages don’t settle this, and it’s the thing that decides which one fits your week.

Fathom’s free plan offers a choice between bot-free capture, which the page marks as being in beta, and a meeting bot that joins the call as a participant. That distinction matters more than it sounds. Some companies have a policy about visible recorders in customer meetings. Some people just find the bot rude, which is a preference rather than a policy, and it still drives the purchase.

Granola runs as a desktop app and writes notes against the meeting you’re sitting in. I didn’t test its capture path for this post, so I’m not going to describe mechanics I haven’t verified myself. Its pricing page does mention MCP integration and API access on the Business tier, which tells you where the product is heading even if it says nothing about how it listens.

Plaud sits outside the computer entirely. The device records, you sync it, the plan transcribes. No meeting invite, no participant list, no software on the call. That is a different product category wearing fairly similar marketing.

Put plainly: if your important conversations happen inside Zoom, Meet or Teams and you want them summarised afterwards, buying hardware is a strange way to solve that problem. If half of them happen in rooms, on doorsteps, or in a car park after the meeting, the two software options aren’t in the running.

One more practical difference, and it’s the unglamorous one. Software notetakers only capture the calls you remembered to run them on, and Plaud only captures the conversations you remembered to press a button for. Both fail the same way, for the same very human reason. The difference is that pressing a button on a thing in your pocket is a slightly easier habit to build than remembering to launch an app before a meeting that started ninety seconds ago.

First-year cost, side by side

Arithmetic from the listed figures, nothing clever: $19.99 a month across twelve months is $239.88, and adding the $159.99 sale price of the Note Pro gives $399.87. If the sale ends and the device goes back to $189.00, that same setup is $428.88. Worth knowing before you decide the hardware is a rounding error.

Teams change the shape of this, and the two-seat case is the one people actually ask about. Fathom Team on annual billing is $15 a month per user with a two-user minimum, so two seats run $360 for the year. Granola listed Business at $14 a month without spelling out a seat minimum the way Fathom’s page does, so I’m not going to state a two-person Granola figure I can’t source. Plaud Team on annual billing is $20 per user monthly in the first year, which is $480 for two seats, and you still need two devices on top. At the Note Pro sale price that’s another $319.98, taking a two-person Plaud setup to just under $800 for year one.

Plaud team buyers’ annual per-user rate moves from $20 to $25 once the launch offer ends on 31 August 2026, by the page’s own wording. Two seats at $25 is $600 a year, hardware already paid for. Still the most expensive of the three. Still possibly the right call, if the conversations that matter are happening away from a desk.

A different meter, briefly

The three above charge for minutes, history or seats, and all three are looking backwards at a call that already happened. Craqly charges for the call itself. Its pricing page on 20 August 2026 reads “1 credit = 1 minute of live session · credits reset every month”, with 20 credits on the free Starter tier, 180 credits on Pro at $19 a month, and stealth mode listed on all four consumer tiers rather than held back for the top one. Different job, different meter. Just don’t buy one expecting the other, which is a mistake I have watched people make with notetakers more than once.

What I’d pick, and where I’m still unsure

Ranking Plaud vs Granola vs Fathom by price alone gets you the wrong answer, because the three of them aren’t solving the same problem. Ranking them by where your conversations physically happen works much better.

If you take calls on a laptop and you’ve never paid for a notetaker, start on Fathom’s free plan and find out whether the limited action items actually annoy you. They might not. If they do, $16 a month annually is a small bill. If you live in Notion or HubSpot and you want notes that behave like a working document rather than a transcript, Granola at $14 is the cheaper paid tier of the two, and the integrations list on its Business plan is the reason to pick it. If a real share of your conversations happen away from a screen, Plaud is the one here that can be in the room with you, and you should price it as roughly $260 to $400 for year one rather than as a monthly subscription.

The bit I’m not confident about is transcription accuracy, because I didn’t run the same audio through all three, and a pricing page tells you nothing about that. If accuracy is your deciding factor, my table won’t help you. Our guide to AI meeting notes covers what to test for, and honestly the best move is to push one real thirty-minute call through the free tier of whichever two you’re torn between, then read both outputs the next morning once you’ve forgotten what was said. That’s the test that decides it.