The short version
- Everything we publish is signed by the Recommended by Agents team. You should be able to check our work without having to trust whoever wrote it.
- Every figure comes from data we collected or a source we link to, with the date we checked it.
- The Index data is published under CC BY 4.0, so anyone can rerun our numbers.
- Nobody can pay for a place in the Index or for coverage in our insights. Clients are marked as clients.
- When we get something wrong, we fix it, add a dated note to the page, and list it below.
1. Who signs our work
Articles and Index editions carry the team byline, not an individual's name. That is deliberate. Our claims are about what AI agents do, and those claims should stand on evidence anyone can inspect: the questions we asked, the answers we coded, the method we followed and the sources we cite. We publish all four.
If you want to challenge a finding, write to hello@recommendedbyagents.com. You will get an answer from the people who did the work.
2. How we research
- First-hand where we can. The Index and our store teardowns come from our own collection: questions put to the agents, and pages run through our readiness check. Each one says when it was collected.
- Sources linked. Anything we did not measure ourselves links to where it came from: vendor documentation, announcements, standards and other research.
- Claims labeled as claims. When a company says its agent does something we have not tested, we say it is the company's claim. Predictions are labeled as predictions.
- Limits stated. Each piece gives its sample size and what it cannot tell you. Edition one of the Index, for example, asked each question once per agent, and the page says so.
- Dated. Agents change quickly. Every article shows when it was published, and when it was last updated if it has been.
3. How we use AI
We use AI tools in our work, and we say where:
- Index coding. Coding passes for the Index can be AI-assisted, working against our written codebook. Each edition says which passes were AI and how disagreements were settled. The coded data is published so the results can be checked.
- Writing. We use AI tools when drafting. Every figure, quote and claim is checked against its source before we publish.
- Images. Images on insights articles are AI-generated illustrations. They do not show real people, products or events.
4. Independence
- Nobody can pay for a position in the Index, and nobody sees results before they are published.
- Our clients are ranked the same way as everyone else. A client that appears in an Index category is marked as a client.
- When an insights article names a client, it says so.
- We do not accept payment, free products or other benefits in return for coverage.
The rules for the Index are set out in full in the methodology.
5. Corrections
If you think something we published is wrong, email hello@recommendedbyagents.com with the page and what you think is wrong. We read every report.
If we were wrong, we correct the page, add a dated note at the end of it saying what changed, and list the correction below. If a correction changes an Index result, we update the data files and the embed too. Typos and broken links are fixed without a note.
6. Corrections log
No corrections so far.