TheShortlist Report

Measured research on who gets recommendedLatest reading

Counts, not opinions · No paid placement · Method published

About

One researcher, one method, and a number above every table telling you how much data it rests on.

Why this exists

Buyers have started asking AI assistants which agency to hire, and the assistants answer. What they answer with is whatever has been written about those agencies online, which in this category is mostly the agencies’ own marketing sites. There is no independent count of who actually gets recommended, so we made one.

The method is dull on purpose. Ask the question a buyer would ask. Ask it many times, because the answer moves. Write down every firm named. Publish the tally, the sample size and the date. Anyone with access to the same assistant can run it again and check.

Who writes this

Reports are bylined to The Shortlist Report rather than to an individual. The rankings are produced by a measurement pipeline and an editorial process, not by a writer with a view, and putting a person’s name on a count they did not personally make would be the more flattering choice rather than the more accurate one.

That is a deliberate trade. A byline names someone you can hold responsible, and we have moved that accountability to two pages you can check instead: who owns this and how it is funded, and exactly how every number is produced. If a page here is wrong, the publication is answerable for it and there is an address to write to.

What the measurement runs on

Prompts are issued to AI assistants on a schedule, and every response is stored whole rather than summarised, so a count can be re-derived later from the original text instead of trusted from a spreadsheet.

Across the 91 questions published so far that comes to 16,604 stored responses naming 509 firms. Every report links the counts it drew from those responses as a CSV and a JSON file, so you can check our arithmetic without asking us for anything.

What we cover, and what is next

We currently publish on managed IT service providers, Amazon agencies, fractional CFO and outsourced accounting firms, TikTok Shop agencies, Salesforce and CRM implementation partners, AEO, GEO and AI visibility agencies, NetSuite and ERP implementation partners and executive search firms. TikTok Shop agencies came first, because the category is new enough that buyers have no established way to compare firms and old enough that plenty of firms exist. Each category since has had to clear the same bar: a question gets a page only when there is a measured sample behind it, and a category only earns a place here when the sample can actually tell firms apart. Where four firms sit tied at the top of every response, we have measured it and not published it — a ranking that cannot separate anyone is not a ranking. No data, no page.

What we collect from you

Google Analytics, counting visits, and nothing else. No account, no newsletter, no advertising, no retargeting, and nothing sold or shared — including with the firms we rank, who are not told they have a page here, let alone who read it. Every page works with JavaScript switched off, so blocking it costs you nothing. The full detail, including how to stop it.

A publication that asks firms to accept being measured in public should be willing to say what it measures about its own readers.

Corrections and contact

Errors, misnamed firms, and firms merged wrongly: how corrections work. Everything else — including requests to be removed, which we will explain rather than action — goes to the same place.