How we rank agencies

Every agency here is assessed against the same five criteria, from the public record. This page sets out what those criteria are, how the rankings are produced, and what our review does not cover, because a ranking you cannot interrogate is not worth reading.

The five criteria

1. Demonstrated client work

   We look for work we can verify from published sources: named clients, described outcomes, case studies with numbers attached. Claims an agency makes about itself are recorded as claims, and you will see them attributed that way in our listings rather than stated as fact.

2. Technical depth

   Whether the agency can build the thing or only assemble it. We look at what they publish about their own methods, the tooling they name, and whether they can describe how something is tested and what happens when it breaks.

3. Specialization fit

   A firm that does one category well outranks a generalist claiming all of them. Our listings say who each agency is for, which is only useful if the answer is narrow. Where a large firm treats a category as one line on a wide menu, we say so.

4. Pricing transparency

   Agencies that state real numbers rank above agencies that will not. Most of what looks like pricing in this market is a directory minimum rather than a published rate, and we distinguish the two.

5. Public accountability

   Whether an agency can be checked from the outside: a verifiable location, a stated team size, a founding year, an independent review record. Several agencies in this directory publish none of these, and we show that rather than hide it.

How reviews are conducted, and their limits

Every listing is desk research. We read what an agency publishes about itself, what third-party directories record about it, and we cross-check the two against each other. Where they disagree, which happens often, the listing says so.

Some things we do not currently do, and it is better that you know which. We have not contacted these agencies as a prospective client, so we cannot tell you how quickly they reply or what they ask before quoting. We have not conducted reference calls with their clients. We have not audited any performance figure an agency publishes about itself. Where a listing repeats such a figure, it is attributed to the agency rather than stated as fact.

The date on each listing is the date we last reviewed the public record for that agency. Our intention is to re-review quarterly, and the date will show you whether we have.

This is a directory built from the public record. That is a real limit, and an agency that looks strong here may still be wrong for you. The matching service exists partly because a page cannot answer that question.

Ratings from other directories

Agencies can ask to be listed. Every request and its outcome is recorded, anonymized, on the stats page; most are declined, and that page explains the bar.

Some listings show a rating from Clutch, DesignRush, G2, Google, or the n8n or Airtable partner directories. Those two directories are run by the vendor whose partners they list and anyone can post to them, so their ratings are shown on a listing and carry no weight in the ranking. monday.com's partner directory is treated the same way: its satisfaction rating and project count are quoted in the text of a listing and carry no weight. Those are cited as facts about what another site publishes, with a link so you can check them yourself and the month we recorded them. We do not reproduce the reviews behind those scores, because they belong to the site that collected them.

We do not average those scores into a rating of our own, and you will not find a number out of five attached to the AI Agency Radar name anywhere on this site. A score we did not compute, presented as ours, would tell you nothing about our review and quite a lot about our willingness to borrow someone else's credibility. Our assessment is the written verdict, the pros and the cons. Read those instead.

A rating from elsewhere never affects rank order here.

Independence

Any agency can be added for free, and no agency can pay for a position in a ranking. Where a paid placement exists in future it will be labelled as such and will never affect rank order. Activity on the Lead Exchange, as sharer or executor, has no effect on any ranking, badge or listing, and ranking position is not an input to lead matching.

Every listing carries at least one drawback. This is a requirement of the format, not an editorial mood: a directory where every entry is excellent is a directory that is not telling you anything.

How the score is calculated

Ranks are produced by a formula, not by hand. Every number below is read directly from the file the scorer uses, so what you are reading is the formula that actually ran, not a description of it written separately and left to drift.

An agency is scored only on the components it has evidence for, and the result is normalised by the weight of those components. That matters more than it sounds: most agencies in this directory carry no third-party rating and most hold no certification, and treating an absence as a zero would have ranked agencies by whether they maintain a directory presence rather than by their work.

Third-party ratings

Ratings are weighted by how many reviews stand behind them and by how easily each source can be influenced. A score from a handful of reviews is pulled toward our own assessment rather than taken at face value.

Worth knowing before you read too much into any score, ours included: across this directory the median third-party rating is 4.9, and 92 percent sit at or above 4.5. On these platforms a 4.5 is not a strong result, it is a below-average one, and a raw number out of five separates agencies far less than it appears to.

The weight of a source applies to its review count as well as its score. Since version 1.2.0, which took effect on 9 October 2026, ten Google reviews carry the confidence of three on Clutch; before that the raw count fed the confidence term, and an agency reviewed only on Google reached the same confidence as one with the same number of verified reviews.

Version 1.2.1, from 21 October 2026, corrects how price transparency is read. A listing that said no price is published on the agency's own site was being scored as if one were, because the sentence mentions the site; such a listing now scores as a directory figure. Twenty-seven listings moved down on that criterion.

Certifications

Credentials can raise an agency's score and never lower it. Before version 1.1.0, which took effect on 2 October 2026, a single modest partner tier was averaged in at its own low value and could leave an agency below where it stood with no credential at all; the component now counts only where it lifts the result.

Only credentials assessed by someone other than the agency count. Anything self-awarded is worth nothing here, and a partner tier we could not verify against the vendor's own directory is excluded rather than taken on the agency's word. Credentials stop adding to the score after a few, so collecting badges does not move an agency up.

A tier in Anthropic's Claude Partner Network counts as a credential once we have found the firm at that tier in Anthropic's own directory at partnerhub.claude.com: Preferred carries more than Select, because Anthropic grants it to partners that have gone further with it. Anthropic has not published what each tier requires, so a tier is weighed as one credential among others and the Claude Partners ranking rests on the Claude work each firm publishes.

The two floors

A rating below 3.5, on enough reviews to mean something, holds an agency below rank 10 however well it scores elsewhere. So does a specialization score below 2, which is our judgement that a firm is only marginally in the category it is listed under.

Both exist for the same reason. A reader who sees an obviously wrong name near the top of a ranking stops trusting the ranking, and they are right to.

The Top choice badge

Given to the top three of any ranking we publish, which covers the full category rankings and the ones filtered to a single industry. An agency sitting sixth in n8n Agencies can be first in n8n Agencies for SaaS, and its badge names the ranking the position came from.

First place carries gold, second silver, third bronze, and every other listed agency carries the green ranked badge. The colour follows the position and nothing else.

The Top choice designation is a separate judgement and it is withheld when the evidence does not support it: a rating below 4.3 on the same card, or a specialization score under 3. An agency held back that way keeps the position its score earned and the badge for it, and loses the designation on our ranking tables. An agency with no reviews at all can earn the designation on the strength of what we could verify ourselves.

When it recalculates

Ratings are refetched and every category re-ranked every two weeks. A rank can therefore move without any editorial decision, because an agency gained reviews or lost them. The formula version and the date it took effect are recorded with each run.

Corrections

If a listing is wrong, write to hello@aiagencyradar.com and we will check it. We correct factual errors on any listing, including our own assessments where the evidence supports it.