PUBLISHED RESEARCH

Published research.

We give a ghost user a job to be done in a product it has never used, no script and no hints, and publish what happens. Same personas, same jobs, every time — so each run is comparable to the last one, and to everyone else's.

READ AGAINST EACH OTHER
  1. PUBLISHED SEPTEMBER 2026 · THREE PRODUCTS, ONE BUYER

    We gave one ghost user the same job in Linear, Notion AI and Airtable: work out what it costs and whether to adopt it, before signing up.

    One ghost user, one job, three products. All three of Linear, Notion AI and Airtable answered what they do, what they cost for five people, and what they want at the door — from public pages, without an account. Where they differ is where the answers live.

    Linear · Notion AI · Airtable · desktop · 1440px →
READ WITH NOBODY IN THE LOOP
  1. PUBLISHED OCTOBER 2026 · FOUR PRICING PAGES, NOBODY IN THE LOOP

    We read four pricing pages with nobody in the loop: Linear, Notion AI, Airtable and Retool. Retool shows its first paid price 1 character from the words saying what that price is per. Notion AI takes 315.

    Four pricing pages — Linear, Notion AI, Airtable and Retool — read with no buyer and no model in the loop. How far each page puts its first paid price from the words saying what that price is per, and how much of what it costs lives on a site of its own.

    Airtable · Retool · Linear · Notion AI · no buyer, no model →
  2. PUBLISHED OCTOBER 2026 · FIVE PRICING PAGES, NOBODY IN THE LOOP

    We read five pricing pages with nobody in the loop: Linear, Notion AI, Airtable, Retool and monday.com. The middle page shows its first paid price 19 characters from the words saying what that price is per. Notion AI takes 315.

    Five pricing pages — Linear, Notion AI, Airtable, Retool and monday.com — read with no buyer and no model in the loop. The middle reading is 19 characters between a page’s first paid price and the words saying what that price is per.

    monday.com · Airtable · Retool · Linear · Notion AI · no buyer, no model →
  3. PUBLISHED OCTOBER 2026 · FIVE PRICING PAGES, READ A THIRD TIME

    We read the same five pricing pages a third time. Where our count is not the same as last time, it is on a page that does not give the same count on two reads of itself — so this page says what it read, and not whose doing it was.

    These five pages — Linear, Notion AI, Airtable, Retool and monday.com — read a third time with nobody in the loop. What the readings have in common, what they do not, and why this page does not say whose doing either difference was.

    monday.com · Airtable · Retool · Linear · Notion AI · no buyer, no model →
  4. PUBLISHED OCTOBER 2026 · FIVE PRICING PAGES, READ AGAIN THE FOLLOWING WEEK

    We read the same five pricing pages again the following week. Where a page gained, lost or renamed a control, this page names the control — rather than counting the edits, because on one of these pages the count is the part that misleads.

    The same five pricing pages — Linear, Notion AI, Airtable, Retool and monday.com — read again the following week with nobody in the loop. Which controls each page gained, lost or renamed, named one at a time rather than counted.

    monday.com · Airtable · Retool · Linear · Notion AI · no buyer, no model →
  5. PUBLISHED OCTOBER 2026 · FIVE PRICING PAGES, READ AND READ AGAIN SINCE THE SECOND OF OCTOBER

    Of five pricing pages we have been reading since the second of October, Linear’s and monday.com’s are the ones whose control counts are still the numbers they were. The other three we will not say that about, and the reason is different for each of them — including one whose count is not the number it was when we started, which is how a reader can tell that the instrument behind this page is able to see a difference at all.

    Five pricing pages — Linear, Notion AI, Airtable, Retool and monday.com — read and read again since the second of October. Linear’s control count has not moved across 40 readings and 519 hours of watching. monday.com’s has not moved across 9 readings and 144 hours. Airtable’s changed this week, and the remaining pages we will not characterise in either direction, with the reason given for each.

    monday.com · Airtable · Retool · Linear · Notion AI · no buyer, no model →
  6. PUBLISHED OCTOBER 2026 · FIVE PRICING PAGES, FETCHED AGAIN AND AGAIN RATHER THAN ONCE

    Two of these five pricing pages are not one page. Retool’s hands out more than one count of itself, reading after reading, and the controls our fetches disagree about are the same ones every time — Contact sales, Book a demo and Request trial. Our own fetches of Notion’s page disagree with each other too, though far more rarely. Linear’s, Airtable’s and monday.com’s pages have handed the same count to every fetch of them, on every reading we have ever taken — and that is the half that makes the first sentence worth reading rather than a complaint about our own instrument.

    Five pricing pages — Linear, Notion AI, Airtable, Retool and monday.com — fetched several times a sitting, over and over, since the middle of October. Two of them do not hand the same control count to every fetch of them: Retool’s on seven readings of it, Notion’s on two readings of it. The other three pages — Linear’s, Airtable’s and monday.com’s — have handed the same count to every fetch of them on every reading we have taken.

    monday.com · Airtable · Retool · Linear · Notion AI · no buyer, no model →
ONE PRODUCT AT A TIME
  1. PUBLISHED SEPTEMBER 2026

    We gave a ghost user one job in Linear: work out what it costs and whether to adopt it, before signing up.

    The front page. It says 'The product development system for teams and agents' and everything under it is agents drafting PRDs and opening pull requests and an Opus 5 panel doing triage. I lead five people who track their work in a shared spreadsheet, and I could not tell from that page whether this was built for me or for a team two years ahead of me. If the price had not been one click away in the nav I would have closed the tab and gone to the next tool on my shortlist — not because it looked bad, but because it looked like it was for somebody else.

    completed in 6m 59s · desktop · 1440px →
  2. PUBLISHED SEPTEMBER 2026OUR OWN PRODUCT

    We gave a ghost user the same job in GhostUser, eight days later: find out where its own signup loses people.

    Reading the FAQ's own pricing answer and finding there is no number anywhere on the site. 'The first run is free and there is no card up front... After that we quote it rather than publish it.' No range, no per-run price, no 'from £X'. The footer link labelled Pricing is not a pricing page at all, it is an anchor into the FAQ. That is the one question I came in needing for Thursday, and the product's answer is that I can only have it by giving them my name and work email and waiting overnight. Everything else I could work out for myself in four minutes; the number is the one thing they hold back, and it turns a self-serve evaluation into a lead-gen conversation at the exact moment I was ready to move.

    partially completed in 5m 03s · desktop · 1440px →
  3. PUBLISHED SEPTEMBER 2026OUR OWN PRODUCT

    We gave a ghost user one job in GhostUser: find out where its own signup loses people.

    Typing my product URL into 'Release a ghost user' — the button sold with 'First run's free, no card' and 'No integration, no setup, no permission' — and landing on /contact: 'Talk to a human. Pilots are hands-on right now.' Name, work email, optional LinkedIn, a topic dropdown, product URL, message, Send, 'We reply within a day.' Then the line above it, 'Prefer to skip the call? A ghost user doesn't need one. Add your product URL →', turned out to be an anchor to the same form on the same page. I tapped the escape hatch and it scrolled me to the thing I was escaping. That is where I stopped believing the home page, and it is also where my one unanswered question — what does this cost — became something I can only get by identifying myself and waiting overnight.

    partially completed in 5m 13s · phone · 390px →
  4. PUBLISHED SEPTEMBER 2026OUR OWN PRODUCT

    We gave a ghost user the same job in GhostUser four days later: find out where its own signup loses people.

    Clicking 'Ask what a run would cost →' from the FAQ and landing on a lead form. The FAQ had just told me the first run is free and that after that 'we quote it rather than publish it' — so the link promising the number instead asked for my name, work email and optional LinkedIn, and promised a reply within a day. That is the one question I came in needing for Thursday, and the product's answer is that I cannot have it without identifying myself and waiting overnight. It was confirmed at the end of the run when 'Release a ghost user' — the free, no-card, no-setup path — dropped me on the same form. Every road on this site ends at one contact form, and I left with no price and no run.

    partially completed in 5m 35s · desktop · 1440px →
  5. PUBLISHED SEPTEMBER 2026OUR OWN PRODUCT

    We gave a ghost user one job in GhostUser: find out where its own signup loses people.

    Looking for what it costs. There is no pricing page, no pricing link in the nav or the footer, and the FAQ answers with 'the first run is free' and then says outright that no numbers appear anywhere on the site. Nothing here can be taken to a budget conversation.

    partially completed in 3m 50s →
THE FORMAT

What a writeup looks like.

Timestamps down one side, screenshots in the middle, and the ghost user's reasoning pinned to the moment it happened. The time it spends stuck is the finding as often as the words are.

EVERY WRITEUP CARRIES
Outcome
Completed, partially completed, or abandoned. Nothing fuzzier.
Time to outcome
How long it took, and where the clock went.
The wall
The single moment the ghost user lost the most ground. One per run, forced choice.
Recommendation
Would it pass the product on — inside its own company, and outside it.
The trace
Every step, timestamped, with the reasoning stated before the action.
What a run needs from you, and what it does once it starts →

Want to see where your product loses people?

Give a ghost user a URL