Read AI Review 2026: Engagement Scores, Pricing, Verdict

Meeting Assistants

Laxis editorial · Updated 10 September 2026 · 10 min read

Read AI is a meeting assistant that records your calls, writes them up, and then does something few of its rivals attempt: it grades the room. One quick disambiguation first, because the phrase pulls two unrelated crowds. This review covers Read.ai the meeting assistant. It has nothing to do with handwriting recognition, cursive, OCR, or reading text aloud. If that's what brought you here, this is the wrong page.

Still with us? Good, because what's left is the most opinionated product in this category. Every other note-taker is trying to answer what was said. Read AI also tries to answer how did that land. It measures sentiment and engagement, folds them into a single Read Score, and hands you a dashboard showing who spoke, who leaned in, and who quietly checked out.

Depending on which side of that dashboard you sit on, that's either the most useful thing anyone has built here or the most uncomfortable. Both readings hold up, so this review takes both seriously — along with what Read does across Zoom, Meet and Teams, where the 2026 product has drifted, what each plan costs, and the complaints worth slowing down for.

What Read AI actually does

The core loop is familiar. Connect a calendar and Read's assistant joins your scheduled meetings on Zoom, Google Meet and Microsoft Teams — platforms Read plugs into rather than competes with — records, transcribes, and produces a Meeting Report: summary, topics, action items, key questions, and a speaker-separated transcript. There are Windows and macOS apps, iOS and Android apps for in-person conversations, and a Chrome extension that pushes the same machinery into Gmail and Google Calendar.

Where Read diverges is the second half of that report. A Deep Dive tab holds talk-time splits, topic timelines, a personal coaching readout covering pace, filler words and something Read labels charisma, and the scores themselves. Paid tiers add video playback and auto-cut highlights, so you can jump to the moment a number moved instead of taking the number's word for it.

The product has pushed well past meetings, too. It summarises email and chat, threads Slack and Teams context into your inbox, and now ships an MCP server, a Claude connector and a ChatGPT app — which tells you plainly where the company thinks this is heading.

The Read Score, and what it is actually measuring

The Read Score combines two sub-scores. Sentiment estimates how positively or negatively attendees are reacting, using natural-language analysis of what's said plus facial expressions, head movements and body language from video. Engagement estimates involvement and attention from those same visual cues plus proportional talk time and vocal features — pitch, volume, intonation. Both run live. Read publishes the scale: above 80 is good, 70 to 80 okay, below 70 poor, with the platform-wide average shown next to yours.

There's an individual layer too: open the Deep Dive tab of any report and you'll find a Participant Score, your own engagement and sentiment plotted against the meeting average and against the average Read user.

Two things here deserve credit rather than suspicion. Users in the EU and UK get a Read Score calculated with no video or facial input at all, which reads as a considered response to European rules rather than a grudging one. And any user, on any plan including free, can switch affective metrics off in account settings — after which the scores stop being computed, not merely hidden.

Read is also specific about a line it won't cross: its documentation says it does not detect whether your meeting window is minimised or hidden behind something else, restricting itself to audio, video and screenshare — what any participant in the call can already observe.

The part that makes people uncomfortable

Now the other side, because it's real and skating past it would be dishonest.

A manager coaching a team gets genuine value from this, and not the vague kind the marketing implies. A rep who talks for seventy-eight percent of a discovery call has a problem they cannot see from inside the call. A recurring meeting where engagement sags every time one agenda item appears is telling you something concrete. Read catches patterns you'd otherwise need a second observer in every meeting to spot, without that observer's instincts about who seems impressive.

The trouble is that the people generating those numbers never applied for the job. A score derived partly from whether you were looking at your camera rewards one style of presence: extroverted, well-lit, at ease on video. It marks down the person taking notes on paper, the colleague on a rural connection, the neurodivergent employee who thinks best while looking away, and anyone dialling in from a car park. None is less engaged. All will score lower.

There's a second-order effect nobody has measured, either. Once a team knows the meeting is graded, meetings change. People perform attention. Disagreement softens, because negative sentiment now appears on a chart with your name beside it. You end up optimising the metric and losing what it stood for.

It's worth saying plainly that scoring people is a design choice, not an inevitability of the category. Plenty of note-takers deliberately stop at the transcript and the follow-through. Laxis is one of them: it captures Zoom, Google Meet and Microsoft Teams calls across 100+ languages and extracts action items with an owner attached to each, but it doesn't rate how enthusiastic anyone appeared while agreeing to them. Neither approach is automatically correct. They're built for different buyers, and working out which buyer you are saves a great deal of arguing later.

Settle this before the trial, not after it. Decide in advance who is allowed to see participant-level scores. Aggregate meeting health, visible to whoever ran the meeting, is a defensible use that most teams shrug at. Individual engagement numbers flowing up to somebody's manager without the employee ever agreeing to it is a different product, and it's the version that generates the furious reviews. Read hands you the switch to disable affective metrics entirely; whether you flip it, and who you tell, is a policy question rather than a software one.

Beyond meetings: email, messaging, and the digital twin

Read's ambition in 2026 isn't really meetings, and that's the shift most buyers miss. The pitch now is a layer across everything you say at work — calls, Gmail, Slack, Teams — with search running on top. Ask Read is the search half: one question, answered from your connected sources, with the Chrome extension surfacing related meetings and messages inside a Gmail thread and offering to draft the reply. It sits on the free plan, which is generous positioning for the software Read most wants you to depend on.

The Digital Twin goes further. Ada, launched in early 2026, handles scheduling in your inbox, proposes times, and fields routine questions from your knowledge base, checking in with you before sending anything beyond a calendar reply. Whether that reads as thrilling or alarming is mostly temperament.

So Read AI is no longer competing only with note-takers. It wants to be the thing that reads all of your work communication — a much larger permission to grant than "take notes in my Monday standup," and one that deserves a conversation with whoever owns security where you work.

Read AI pricing in 2026, plan by plan

Read charges per seat, with 25% off for annual billing. These are the published plans.

Prices and limits taken from Read AI's published plans page, September 2026. Volume discounts start at 100 licences.

Three details the table doesn't shout. Every tier, free included, carries a line reading "30 meetings measured per month" — the analytics that make Read distinctive have a ceiling of their own, separate from the transcript count. Video playback sits on Enterprise, so a Pro subscriber gets the numbers without the footage to check them against. And Read's premium features ship at no extra cost inside several Zoom One plans via Zoom's Essential Apps bundle, worth checking before anyone buys a seat.

The free plan, meanwhile, is a trial wearing a nicer label. Five meetings a month, an hour each, no playback. Anyone with an ordinary calendar clears it inside the first week.

Read the middle of the price list, not the ends. Free tells you nothing about living with the product, and Enterprise+ is a procurement decision rather than a preference. The tier that reveals what Read AI really costs you is Pro at $15 a seat annually — unlimited transcripts, no playback. And if the analytics are the whole reason you want it, budget for Enterprise at $22.50, because a score you can't replay is a score you can't argue with.

Why the reviews split so hard

Read AI's reputation is unusually split, and the split is itself the finding. G2 has it around 4.0 across a fairly small review count. Trustpilot — where people go when they're angry rather than when procurement asks for a reference — sits dramatically lower, dominated by one-star entries.

The recurring grievance isn't about output quality. It's about consent and control. Users describe Read's assistant appearing in calls they never added it to, arriving through a colleague's calendar connection and proving stubborn to remove — Read maintains a support article on how to stop it joining meetings, which tells you how often that question lands. Several reviewers report difficulty deleting recorded data afterwards.

It escalated past reviews, too. At least two US universities publicly restricted or prohibited Read AI on institutional accounts during 2025, citing data and privacy risk. That's a serious signal — not proof of mishandled data, but proof that security teams examined how the product propagates and disliked what they found.

In fairness, much of this traces to a viral growth design that shares reports outward and pulls colleagues in automatically — a distribution decision rather than a data-handling one. Read has since tightened admin controls, added domain capture on the top tier, and published a privacy centre. Deploy it deliberately, with a workspace and an administrator, and the experience is very different from meeting it in a call you didn't schedule.

Where Read AI fits, and what to weigh it against

Best for: teams that genuinely want meeting analytics and will use them for coaching rather than surveillance. An enablement lead or a sales manager running a team that has been told clearly what gets measured, and has agreed to it, will get more from Read than from anything else here. It also suits someone who wants one assistant spanning meetings, email and chat and is comfortable granting that reach.

It's a poor pick if what you want is a quiet note-taker nobody has to have feelings about, if your meetings run across languages, or if "no third party in the call" is written into a client agreement.

On that last group: Laxis is an AI meeting assistant built around the follow-through instead of the analytics. Bot-free capture is one of its standard modes, recording and summaries run across 100+ languages with live side-by-side translation appearing as people speak, action items come out with owners attached, and the write-up lands on the right HubSpot or Salesforce record. Pricing is flat: free with 300 transcription minutes a month, Premium at $15.99 per user, Business at $29.99, both cheaper billed annually, with an AI voice keyboard included.

The honest limits, since nothing here is exempt. Laxis runs its processing on servers rather than locally on your laptop, so a policy insisting audio never leaves the device rules it out at the door. The free tier stops at 300 transcription minutes a month — plenty to evaluate with, not enough to run a full calendar on indefinitely. And it's a smaller brand than the incumbents, which shows up mainly as fewer third-party walkthroughs when you're troubleshooting something obscure.

If you'd rather survey the field than make a straight swap, our ranking of AI note-takers lays out what each one is genuinely good at. For the tool with the deepest integration list, our roundup of Fireflies.ai alternatives walks through that trade. If multilingual calls are the reason you're shopping at all, we looked at how well real-time meeting translation actually works on its own. And for the architectural opposite of Read — an extension that never joins the call — our Tactiq review covers what that buys and what it costs.

Notes that move the work forward, without grading the room

Laxis writes up your Zoom, Meet and Teams calls in 100+ languages, translates live while people are still talking, hands back action items with names attached, and files the result in your CRM. No participant score included.

Get started with Laxis

The verdict, in two parts

Read AI is the most opinionated product here, and opinionated is not a synonym for wrong. The scoring rests on stated methodology, it's disclosed rather than buried, it can be switched off completely, and it gives a coach something a transcript never will. The teams that want it really do want it.

The risk isn't the technology. It's the deployment. Read spreads through organisations quickly and quietly, and the gap between "I chose this" and "this turned up in my call" is exactly where the one-star reviews live. Buy it on purpose, publish what gets measured, decide who sees participant scores, and it's a defensible tool with a real edge. Let it arrive by itself and you'll be having a much less pleasant conversation with your security team.

Frequently asked questions

What is Read AI?

Read AI is an AI meeting assistant that joins calls on Zoom, Google Meet, and Microsoft Teams to record, transcribe, and summarize them, then adds analytics most rivals skip: talk time, sentiment, engagement, and a combined Read Score. It also summarizes email and messages, and offers search across all of it.

How much does Read AI cost in 2026?

Read AI costs $0 for the free plan, $15 per user per month for Pro, $22.50 for Enterprise, and $29.75 for Enterprise+, all billed annually. Monthly rates run 25 percent higher at $19.75, $29.75, and $39.75. Enterprise+ needs at least five licenses, and volume discounts begin at 100 licenses.

Is Read AI free?

There is a free plan, but it is small. Read AI's free tier covers 5 meeting transcripts per month with a 1 hour cap on each meeting, plus summaries, search, the meeting coach, and the mobile and desktop apps. Video playback is excluded. Most people with a normal calendar exhaust it in the first week.

What is the Read Score and how is it calculated?

The Read Score is an average of two sub-scores, sentiment and engagement. Sentiment reads facial expressions, head movements, and the words themselves to judge how positively people are reacting. Engagement adds proportional talk time and vocal features such as pitch and volume. Above 80 counts as good, 70 to 80 as okay, and below 70 as poor.

Can you turn off Read AI's engagement and sentiment scoring?

Yes. Any Read AI user on any plan can disable affective metrics in account settings, and once they are switched off the sentiment, engagement, and Read Score numbers stop being calculated rather than merely hidden. Users in the EU and UK also get a Read Score computed without any video or facial analysis by default.

Why do people complain about Read AI joining meetings uninvited?

Because Read AI spreads through calendar connections and shared reports, its assistant can turn up in calls the host never added it to, usually through a colleague's account. Read publishes a support article on how to stop it joining, and at least two US universities restricted the tool on institutional accounts in 2025 over data and privacy concerns.