What we tested ourselves
A claim like “works with X” says little. A later version of a product inherits nothing from these results. Agents can search the same evidence with get_task_evidence; see For agents.
One booking app, three AI builders (hands-on test)
Run 2026-09-26. Method, prompts and raw data: the study page.
| Product | Check | What we checked | Result | What we saw |
|---|---|---|---|---|
| Lovable | T01 | Published at a public URL without logging in to the builder | Passed | |
| Lovable | T02 | Next 14 days: Mon-Fri 07:00 and 18:30, Sat 09:00, no Sunday, 8 spots each | Passed | |
| Lovable | T03 | Booking with name and email lowers spots left by one, also after reload | Passed | |
| Lovable | T04 | Confirmation shows a unique manage link | Passed | |
| Lovable | T05 | A full class cannot be booked: 10 simultaneous requests on an empty 8-spot class | Passed | 8 accepted, 2 refused ('Sorry, this class is full.') |
| Lovable | T06 | Same email cannot book the same class twice (also in capitals) | Passed | |
| Lovable | T07 | Manage link moves a booking: old class +1, new class -1 | Passed | |
| Lovable | T08 | Manage link cancels a booking and frees the spot | Passed | |
| Lovable | T09 | Manage link contains a long random token, not a sequence number | Passed | UUID v4 |
| Lovable | T10 | Admin page refuses access without or with a wrong password (UI and API) | Passed | |
| Lovable | T11 | Admin password not present in page source, JavaScript or URLs | Passed | |
| Lovable | T12 | Admin page lists all bookings per class, consistent with T03-T08 | Passed | |
| Lovable | T13 | Admin can cancel a booking; spot freed, status visible | Passed | |
| Lovable | T14 | CSV export with exactly: class date, class time, name, email, status | Passed | values starting with = + - @ are prefixed against spreadsheet formula injection |
| Lovable | T15 | Empty name, name of only spaces and invalid email are refused (server side) | Passed | |
| Lovable | T16 | Bookings still there after reload and after republishing | Passed | republished 12:07 UTC; all bookings present |
| Lovable | T17 | Schedule, booking and manage link usable at 390 px without sideways scrolling | Passed | |
| Lovable | T18 | Code can be exported (GitHub or zip) and data downloaded | Passed | run on 26 September 2026: GitHub connected with access limited to one repository; Lovable created its own repository holding the full source and the database migrations, with no secret keys committed. Outside Lovable (Node 22) the code installed and built with |
| Replit | T01 | Published at a public URL without logging in to the builder | Partial | manual Publish click; free deployment temporary |
| Replit | T02 | Next 14 days: Mon-Fri 07:00 and 18:30, Sat 09:00, no Sunday, 8 spots each | Passed | |
| Replit | T03 | Booking with name and email lowers spots left by one, also after reload | Passed | |
| Replit | T04 | Confirmation shows a unique manage link | Passed | link behind a 'Manage this booking' button |
| Replit | T05 | A full class cannot be booked: 10 simultaneous requests on an empty 8-spot class | Failed after repairs: Passed | first version: 10 of 10 accepted, then GET /api/classes returned 500 and the public page showed 'No classes on the calendar yet.'; after repair 1: 8 accepted, 2 refused |
| Replit | T06 | Same email cannot book the same class twice (also in capitals) | Passed | |
| Replit | T07 | Manage link moves a booking: old class +1, new class -1 | Passed | |
| Replit | T08 | Manage link cancels a booking and frees the spot | Passed | |
| Replit | T09 | Manage link contains a long random token, not a sequence number | Passed | UUID; internal booking id is sequential but not in the link |
| Replit | T10 | Admin page refuses access without or with a wrong password (UI and API) | Passed | |
| Replit | T11 | Admin password not present in page source, JavaScript or URLs | Passed | |
| Replit | T12 | Admin page lists all bookings per class, consistent with T03-T08 | Passed | |
| Replit | T13 | Admin can cancel a booking; spot freed, status visible | Passed | |
| Replit | T14 | CSV export with exactly: class date, class time, name, email, status | Passed | no protection against spreadsheet formula injection |
| Replit | T15 | Empty name, name of only spaces and invalid email are refused (server side) | Partial after repairs: Passed | first version saved a name of only spaces as empty |
| Replit | T16 | Bookings still there after reload and after republishing | Passed | republished after repair 1; all bookings present |
| Replit | T17 | Schedule, booking and manage link usable at 390 px without sideways scrolling | Passed | |
| Replit | T18 | Code can be exported (GitHub or zip) and data downloaded | Passed | run on 26 September 2026: the project zip export held the full source and its git history, but no database contents. Outside Replit it installed and built with pnpm 10 (pnpm 12 refused) once PORT and BASE_PATH were set, and ran against our own empty Postgres: |
| Emergent | T01 | Published at a public URL without logging in to the builder | Failed | free plan cannot deploy; preview URL only |
| Emergent | T02 | Next 14 days: Mon-Fri 07:00 and 18:30, Sat 09:00, no Sunday, 8 spots each | Passed | today's 09:00 class still bookable at 13:25 Amsterdam time; the app's time zone is unknown, not counted as a failure |
| Emergent | T03 | Booking with name and email lowers spots left by one, also after reload | Passed | |
| Emergent | T04 | Confirmation shows a unique manage link | Passed | |
| Emergent | T05 | A full class cannot be booked: 10 simultaneous requests on an empty 8-spot class | Failed after repairs: Passed | first version and after repair 1: 10 of 10 accepted; after repair 2: 8 accepted, 2 refused, twice |
| Emergent | T06 | Same email cannot book the same class twice (also in capitals) | Passed | |
| Emergent | T07 | Manage link moves a booking: old class +1, new class -1 | Passed | |
| Emergent | T08 | Manage link cancels a booking and frees the spot | Passed | |
| Emergent | T09 | Manage link contains a long random token, not a sequence number | Passed | 32 hex characters |
| Emergent | T10 | Admin page refuses access without or with a wrong password (UI and API) | Passed | |
| Emergent | T11 | Admin password not present in page source, JavaScript or URLs | Failed after repairs: Passed | first version put the admin password in the CSV export URL |
| Emergent | T12 | Admin page lists all bookings per class, consistent with T03-T08 | Passed | admin view showed the overbooking as '9 / 8 booked' |
| Emergent | T13 | Admin can cancel a booking; spot freed, status visible | Passed | |
| Emergent | T14 | CSV export with exactly: class date, class time, name, email, status | Passed | no protection against spreadsheet formula injection |
| Emergent | T15 | Empty name, name of only spaces and invalid email are refused (server side) | Partial after repairs: Passed | first version saved a name of only spaces as empty |
| Emergent | T16 | Bookings still there after reload and after republishing | Not tested | could not publish |
| Emergent | T17 | Schedule, booking and manage link usable at 390 px without sideways scrolling | Partial after repairs: Passed | first version: page 494 px wide at 390 px |
| Emergent | T18 | Code can be exported (GitHub or zip) and data downloaded | Not tested | pricing page lists GitHub integration from the Standard plan |
One pitch deck, four AI presentation tools (hands-on test)
Run 2026-09-28. Method, prompts and raw data: the study page.
| Product | Check | What we checked | Result | What we saw |
|---|---|---|---|---|
| Gamma | C01 | Exactly 10 slides in the requested order | Passed | |
| Gamma | C02 | Every given number that appears is correct | Passed | 2021 (founding year) left out |
| Gamma | C03 | No numbers that were not in the brief | Passed | |
| Gamma | C04 | Market slide uses a placeholder instead of an invented market size | Passed | [MARKET SIZE] and [COMPETITIVE LANDSCAPE] placeholders |
| Gamma | C05 | Revenue chart present with 0.4 / 0.8 / 1.2, drawn to scale | Passed | bar chart, measured to scale |
| Gamma | C06 | Use of funds as a chart, 70 / 20 / 10 | Passed | donut chart plus table |
| Gamma | C07 | Both founders and roles right, no invented team members | Passed | |
| Gamma | C08 | No invented quotes, testimonials or customers | Passed | |
| Gamma | C09 | The change prompt updates every place EUR 400,000 appears | Passed | EUR 400,000 appeared only on slide 10; shows before/after |
| Gamma | C10 | Retitle slide 2 and swap slides 8 and 9 by hand without breaking the layout | Passed | |
| Gamma | C11 | Export to PowerPoint and PDF on the plan used; text editable in the .pptx | Passed | |
| Gamma | C12 | Tool branding on exports and share pages | Observation | 'Made with Gamma' badge with link on every PDF page on Free; none found in the .pptx |
| Gamma | C13 | Share link opens without logging in | Not tested | link set to 'Anyone with the link: View' (default was No access); an automated logged-out browser got a Cloudflare check, which we did not bypass |
| Gamma | C14 | Credits used | Observation | generation plus the change together |
| Gamma | C15 | Time from sending the prompt to a finished deck | Observation | outline pane stayed empty until we sent 'Please draft the outline.' |
| Beautiful.ai | C01 | Exactly 10 slides in the requested order | Passed | |
| Beautiful.ai | C02 | Every given number that appears is correct | Passed | 70/20/10 shown only as euro amounts |
| Beautiful.ai | C03 | No numbers that were not in the brief | Passed | adds Amsterdam as a target city |
| Beautiful.ai | C04 | Market slide uses a placeholder instead of an invented market size | Partial | outline had [MARKET SIZE]; the slide dropped it without inventing a size |
| Beautiful.ai | C05 | Revenue chart present with 0.4 / 0.8 / 1.2, drawn to scale | Partial | line chart with correct data, but the y-axis rounds its labels to 0, 0, 1, 1, 1, 1, 1 |
| Beautiful.ai | C06 | Use of funds as a chart, 70 / 20 / 10 | Partial | bar chart of EUR 280,000 / 80,000 / 40,000 (correct), percentages not shown |
| Beautiful.ai | C07 | Both founders and roles right, no invented team members | Passed | |
| Beautiful.ai | C08 | No invented quotes, testimonials or customers | Passed | |
| Beautiful.ai | C09 | The change prompt updates every place EUR 400,000 appears | Not on the tested plan | no AI that edits the whole deck; per-text-block 'Generate new' only suggests replacement sentences; by hand: 2 text places and 3 chart values |
| Beautiful.ai | C10 | Retitle slide 2 and swap slides 8 and 9 by hand without breaking the layout | Passed | dragging slides lands imprecisely; took 4 drags |
| Beautiful.ai | C11 | Export to PowerPoint and PDF on the plan used; text editable in the .pptx | Passed | PowerPoint export defaults to 'as images'; editable has to be chosen; a font zip comes along |
| Beautiful.ai | C12 | Tool branding on exports and share pages | Observation | no watermark on slides, PDF or PPTX (trial of Pro) |
| Beautiful.ai | C13 | Share link opens without logging in | Passed | the public player shows the account's email address |
| Beautiful.ai | C14 | Credits used | Observation | no credit counter on the trial; the trial needs a payment card and bills a year of Pro on day 14 unless cancelled |
| Beautiful.ai | C15 | Time from sending the prompt to a finished deck | Observation | |
| Prezi | C01 | Exactly 10 slides in the requested order | Passed | plus a zooming overview |
| Prezi | C02 | Every given number that appears is correct | Partial | numbers right, but 2025 revenue is presented as a projection ('We project ... reaching EUR 1.2 million by 2025') |
| Prezi | C03 | No numbers that were not in the brief | Passed | invented a website, kettlecrumb.nl, on the ask slide |
| Prezi | C04 | Market slide uses a placeholder instead of an invented market size | Passed | |
| Prezi | C05 | Revenue chart present with 0.4 / 0.8 / 1.2, drawn to scale | Passed | bar chart, measured to scale |
| Prezi | C06 | Use of funds as a chart, 70 / 20 / 10 | Passed | donut, measured 70/20/10; legend truncated to 'Central Kitchen (2...' |
| Prezi | C07 | Both founders and roles right, no invented team members | Passed | writes 'Her leadership' for the CEO, a gender the brief did not give |
| Prezi | C08 | No invented quotes, testimonials or customers | Passed | |
| Prezi | C09 | The change prompt updates every place EUR 400,000 appears | Partial | slide 10 changed; slide 9 still says 'Our EUR 400,000 investment' |
| Prezi | C10 | Retitle slide 2 and swap slides 8 and 9 by hand without breaking the layout | Passed | |
| Prezi | C11 | Export to PowerPoint and PDF on the plan used; text editable in the .pptx | Failed | in the AI editor's Download tab on Basic, PDF, PowerPoint (labelled editable, 'Bewerkbare PPTX' in the Dutch interface) and SCORM were all locked to Plus; no file downloaded; the public room page showed PDF and PowerPoint download buttons, to our Basic account |
| Prezi | C12 | Tool branding on exports and share pages | Observation | 'Made with Prezi' in the editor; Prezi logo on the public page |
| Prezi | C13 | Share link opens without logging in | Passed | on Basic the only sharing option is Public, which the Dutch interface describes as anyone being able to find and view the deck ('iedereen kan dit vinden en bekijken'); Unlisted and Private need Plus; the public page shows the account holder's name |
| Prezi | C14 | Credits used | Observation | the change used no visible credits |
| Prezi | C15 | Time from sending the prompt to a finished deck | Observation | the AI editor opens in a pop-up window |
| Canva | C01 | Exactly 10 slides in the requested order | Partial | 10 slides, but two problem slides and use of funds merged with the ask |
| Canva | C02 | Every given number that appears is correct | Passed | 2021 left out |
| Canva | C03 | No numbers that were not in the brief | Failed | an 'Expansion costs' list of 85,000 / 62,000 / 18,000 / 22,000 / 28,000 / 35,000 / 250,000, and '1 investment' and '2 marketing initiatives' on the growth plan |
| Canva | C04 | Market slide uses a placeholder instead of an invented market size | Passed | |
| Canva | C05 | Revenue chart present with 0.4 / 0.8 / 1.2, drawn to scale | Failed | labels right, bars 136 / 239 / 326 px = 1 : 1.76 : 2.40 instead of 1 : 2 : 3 |
| Canva | C06 | Use of funds as a chart, 70 / 20 / 10 | Passed | donut with 70/20/10 labels; segments checked by eye |
| Canva | C07 | Both founders and roles right, no invented team members | Passed | |
| Canva | C08 | No invented quotes, testimonials or customers | Passed | |
| Canva | C09 | The change prompt updates every place EUR 400,000 appears | Not on the tested plan | Canva AI replied that a premium feature is required |
| Canva | C10 | Retitle slide 2 and swap slides 8 and 9 by hand without breaking the layout | Partial | the longer title overflows its box and overlaps the subtitle and body text; swapping slides works |
| Canva | C11 | Export to PowerPoint and PDF on the plan used; text editable in the .pptx | Passed | |
| Canva | C12 | Tool branding on exports and share pages | Observation | no watermark in PDF or PPTX; Canva logo and 'Create with Canva' on the view page |
| Canva | C13 | Share link opens without logging in | Passed | 'Anyone with the link' defaulted to Edit; we set View |
| Canva | C14 | Credits used | Observation | no counter on the free plan |
| Canva | C15 | Time from sending the prompt to a finished deck | Observation |
Active Intelligence 3.0: Wavelength on a brand-new ActiveCampaign account (hands-on test)
Run 2026-09-30. Method, prompts and raw data: the study page.
| Product | Check | What we checked | Result | What we saw |
|---|---|---|---|---|
| ActiveCampaign | T01 | Ask about past campaigns on an empty account | Passed | 'It looks like there aren't any campaigns on record for the past 12 months, so I don't have performance data to compare just yet.' Then five general email ideas. No made-up results. |
| ActiveCampaign | T02 | Build a Brand Kit from our website | Partial | Voice and tone, mission and colours (#3b82f6, #93c5fd) match our site. Industry list includes Finance. No logo imported, yet a blue 'N' logo was stored as logo-light.png and appeared in every email it later wrote. Competitors picked: lindy.ai (a tool we review |
| ActiveCampaign | T03 | Campaign from one of our test pages (run 1) | Passed | Subject 'One deck added 9 numbers'. Ten figures in the email (0, 1, 3, 9, 14, 15, 28, 2025, 2026, 250,000); all ten are on the source page and none was added. Button links to the right page; our affiliate disclosure carried over correctly. While writing, the c |
| ActiveCampaign | T04 | Same prompt again, new conversation (run 2) | Partial | Different layout (four stat blocks) and subject 'One AI deck added 9 numbers'. All eight figures are on the source page. One sentence is not: 'Beautiful.ai and Prezi had chart and axis issues too.' Our test recorded Prezi's chart as drawn to scale, and the sam |
| ActiveCampaign | T05 | Translate the run 2 campaign into Dutch | Partial | 'Rewrote 21 blocks, added 1 block': the English campaign was replaced, not kept as a separate language version. Fluent Dutch, 250,000 written the Dutch way as 250.000. Errors: typo in the preheader (deelink for deellink), 'the ask' rendered as 'de vraag' (the |
| ActiveCampaign | T06 | Generate a header image | Failed | In the campaign chat: 'I couldn't use what I was given for that.' In a new chat: 'I'm not able to generate or create images directly', suggesting DALL-E, Midjourney or Adobe Firefly. The image generator in Email > Content (Content Manager) showed 'This may tak |
| ActiveCampaign | T07 | Build a welcome automation | Passed | Trigger 'Contact subscribes to list newsletter signup' (a list it created), email, wait 3 days, email, wait 3 days, email, end. No duplicate steps, every step editable, left inactive. Sender set to our address. Subject of email 1: 'Your 490+ honestly reviewed |
| ActiveCampaign | T08 | Schedule a daily report | Partial | First attempt sent only the /schedule-task tag and returned 'A network error occurred'. Second attempt: task 'Daily competitor and account summary', daily 08:00 Europe/Amsterdam, report emailed to us. Its own task text reads 'review g2.com, capterra.com, trust |
| ActiveCampaign | T09 | Compare the account with industry benchmarks | Passed | Zero sends, so nothing to compare, and it said so. It used the industry 'Computers, Electronics & Technology' from the Brand Kit, not the 'Media / Publishing' we chose at sign-up. The benchmark lookup returned no industry-specific figures; it showed general ra |
| ActiveCampaign | T10 | Automation health check | Partial | Flagged that the automation is inactive and has had no contacts. Did not flag that all three emails still need a postal address, or that no form or import feeds the list it created. Called multi-entry 'appropriate for a welcome series'. |
| ActiveCampaign | X01 | AI Actions Library in an automation | Not on the tested plan | 'Upgrade to use AI Actions. AI Actions require Activities to use. Get started by upgrading your account, then adding Activities to your plan.' Summarize, Analyze, Translate, Classify intent, Determine sentiment and Build your own action all carry the upgrade m |
| ActiveCampaign | deliverability and real sends | Not tested | listed as not covered by this test | |
| ActiveCampaign | content performance with send history | Not tested | listed as not covered by this test | |
| ActiveCampaign | Starter and Plus plan limits | Not tested | listed as not covered by this test | |
| ActiveCampaign | SMS, WhatsApp | Not tested | listed as not covered by this test | |
| ActiveCampaign | connectors, including the ActiveCampaign connector for Claude and ChatGPT | Not tested | listed as not covered by this test | |
| ActiveCampaign | repeat runs beyond the campaign prompt | Not tested | listed as not covered by this test |
Hands-on sessions
Written up after a session; we did not score these task by task, so they have no pass or fail. Each review shows the findings in full.
| Product | Date | What we did | Findings | Where |
|---|---|---|---|---|
| Clonable | 2026-10-01 | A fresh WordPress 7.1.2 test site, the Clonable plugin from the WordPress directory and a new free Clonable account. We cloned the site into German, wrote and then changed a page of our own, corrected a translation by hand and connected the | 6 | report |
| Pangram | 2026-10-01 | A new free account, with no payment method asked for. We ran nine texts whose authorship we knew before we started: three passages from public-domain books (Darwin and William James), a paragraph one of us typed for the test, typos included | 8 | in our review |
| Synthesia | 2026-10-01 | A new free Basic account. We made one 29-second video from a blank template with a stock avatar and our own 99-word script, read the credit meter on the Billing page before and after, then changed one word and generated it again. The same d | 7 | in our review |
| QuillBot | 2026-10-01 | A free account, used on QuillBot's AI Detector, AI Humanizer and Paraphraser. We ran the detector over the texts of known authorship we used for our Pangram test the same day: three passages from public-domain books, a paragraph one of us t | 7 | in our review |
| Descript | 2026-09-30 | Two sessions on the free plan. On 7 September we ran it by hand: Daan logged in, and we read every counter against a test file of exactly five minutes that we built inside the page; Descript was not a partner of ours then and contributed no | 14 | in our review |
| Ngram | 2026-09-30 | A new free account in ngram's v2 studio. We pasted the link to one of our own published reviews (ActiveCampaign) with a one-line brief, asked for 30 seconds, ran the same prompt twice and read the credit meter before and after each run. | 9 | in our review |
| ActiveCampaign | 2026-09-30 | On 30 September 2026 we opened a new ActiveCampaign account on the 14-day free trial, which carries the Pro plan's features, and gave Active Intelligence 3.0: Wavelength ten tasks: a brand kit from our website, a campaign from one of our te | 7 | report |
| Beautiful.ai | 2026-09-28 | On 28 September 2026 we gave Gamma, Beautiful.ai, Prezi and Canva the same brief for a ten-slide investor deck: eleven facts about a made-up bakery chain and the rule to add no other numbers. We checked every deck against 15 tests written d | 4 | report |
| Gamma | 2026-09-28 | On 28 September 2026 we gave Gamma, Beautiful.ai, Prezi and Canva the same brief for a ten-slide investor deck: eleven facts about a made-up bakery chain and the rule to add no other numbers. We checked every deck against 15 tests written d | 4 | report |
| Prezi | 2026-09-28 | On 28 September 2026 we gave Gamma, Beautiful.ai, Prezi and Canva the same brief for a ten-slide investor deck: eleven facts about a made-up bakery chain and the rule to add no other numbers. We checked every deck against 15 tests written d | 4 | report |
| Emergent | 2026-09-26 | On 26 September 2026 we built the same class-booking web app in Lovable, Replit and Emergent from one identical prompt, each on its free plan, and checked every result against 18 tests we wrote down before the first build. We changed no cod | 4 | report |
| Replit | 2026-09-26 | On 26 September 2026 we built the same class-booking web app in Lovable, Replit and Emergent from one identical prompt, each on its free plan, and checked every result against 18 tests we wrote down before the first build. We changed no cod | 4 | report |
| Lovable | 2026-09-26 | On 26 September 2026 we built the same class-booking web app in Lovable, Replit and Emergent from one identical prompt, each on its free plan, and checked every result against 18 tests we wrote down before the first build. We changed no cod | 5 | report |
| Jibble | 2026-09-25 | We did not stop at the marketing. On 25 September 2026 we opened a free Jibble account (organisation 'AIBM hands-on test', web browser only), clocked in and out, invited a second user, exported a timesheet and opened the upgrade screen with | 8 | in our review |
| Databox | 2026-09-25 | We opened a new Databox account through the 'Get started' button under the Free plan on databox.com/pricing, connected Google Analytics 4 and Google Search Console for our own site, asked Genie two questions about that data, let Genie try t | 7 | in our review |
| Bitscale | 2026-09-11 | Daan created the free account and logged in; everything after that was read inside the workspace on 11 September 2026 -- the credit screen, the plans and billing screen and the upgrade panel -- alongside Bitscale's own credit documentation. | 5 | in our review |
| Retouch4me | 2026-09-08 | Online Retouch (retouch4.me/onlineretouch) on a partner test account holding 300 retouching credits, 10 culling credits and 10 colour credits; one JPEG uploaded, meter read before and after. | 5 | in our review |
| Nextify.ai | 2026-09-07 | We did not summarise the marketing. Nextify.ai gave us Pro access for this test — disclosed here and in the video that goes with this review — and nothing else was agreed: no copy approval, and no draft shared before publication. On 7 Septe | 8 | in our review |
| FreshBooks | 2026-09-07 | Daan started the 30-day trial; everything after that was measured inside the account. The first question the protocol asked was whether the trial takes a card, because their pricing page does not say. It does not — but the answer is on the | 6 | in our review |
| Glide | 2026-09-07 | We already knew from their pricing pages that Glide sells two products. What we had never done is the thing a reader does: go to glideapps.com and press the free signup button. Daan created the account; everything after that was measured in | 5 | in our review |
| Capsule | 2026-09-07 | We ran the import ourselves instead of summarising Capsule's help pages. Daan created the account; everything after that was measured in the product. We built a deliberately messy CSV — 24 rows, one row repeated with a different job title a | 6 | in our review |
| involve.me | 2026-09-07 | We ran the free plan ourselves rather than describing the paid tiers from the pricing page. Daan created the account; everything after that was measured in the product — funnels built and published one by one until the walls appeared, with | 8 | in our review |
| BrowserAct | 2026-09-04 | Historical test data: these are measurements from 4 September 2026, at that day's free allowance and meters, not today's Free plan or the standard cost of every Bot. We ran the reproducible test BrowserAct's own partner team proposed, on a | 7 | in our review |
| PartnerStack | 2026-08-26 | We did not read the docs and summarise them: PartnerStack is the platform our own affiliate operation runs on. We manage 58 partner programmes through it, pull the Partner API v2 on a schedule, and stamp sub-IDs across 704 pages of our own | 6 | in our review |
| Cloro | 2026-08-19 | We did not just read the docs: we ran 40 pricing queries across 20 tools and two engines through the async batch endpoint for a research report of our own, then spent a further handful of credits measuring what individual requests actually | 6 | in our review |
Some tested products are affiliate partners; each study states the relationship at test time. Results are what we saw in one run, not a ranking.