# How We Rank Agencies > The AIAgencies.eu review methodology: who qualifies, what evidence is verified, what Pro buys and never buys, and how freshness works. Canonical page: https://www.aiagencies.eu/methodology Generated: 2026-09-17 · Curated by Gabor Kiss · No paid placements — inclusion, score, and rank are never sold. Let's be honest: most agency directories are "pay-to-play." The agency that bids the highest gets the #1 spot, regardless of whether it can actually ship working AI. AIAgencies.eu exists to fix that for one market: AI agencies and consultancies based in Europe. Since launch we've evaluated 418 agencies and not listed 104 of them. We count the reasons separately, because only one of them is a judgment about the agency: - **49 declined on evidence.** We read the published work and it didn't hold up — no named clients, no verifiable delivery, or claims the site couldn't support. This is the bar, and it's the only number here that reflects it. - **16 out of scope.** Real firms, several of them excellent, that this register doesn't cover: a head office outside Europe, or a product sold to everyone rather than services built per client. Wrong directory, not failed review. - **23 waiting on a city.** Based somewhere we haven't opened yet. Geography is a statement about our map, not about the agency — and this list doubles as our expansion queue. - **16 unreachable.** No working website by the time we looked: dead, parked, or redirected to a company that had acquired them. We can't judge work we can't see. Curation means saying no — a listing here is earned, not bought. It also means being precise about what the no was for. Every agency is reviewed by a named person: **Gabor Kiss**, founder and curator. He works in enterprise software at SAP and applies the discipline used to vet vendors for production systems — read the published work, check the named clients, ignore the pitch. He also runs UXAgencies.com on the same evidence-first model. Here is exactly how the register works. --- ## Who Qualifies for a Listing An agency qualifies when all of the following hold: - **Based in Europe.** Headquarters or a substantive office in the EU/EEA, UK, or Switzerland. A mailbox address doesn't count. - **Verifiable AI delivery.** At least one published engagement describing real AI substance — the problem, the technique, and what the system does — or a demonstrable AI product with an identifiable team behind it. One is the floor, not the goal: how much an agency has shipped, and how deeply it is documented, is what the scores measure. - **Anonymized work counts.** Much of the best AI delivery sits under NDA, and the agencies doing it are often the ones with the strictest clients. A case study that describes the system, the method and the outcome without naming the client is evidence; we do not punish an agency for its client's confidentiality. A named client is stronger and scores higher — but its absence is not disqualifying. - **Evidence, not self-description.** Selling AI services is not the test; showing AI work is. An agency whose published portfolio is websites, apps, or campaigns does not qualify on the strength of an AI page — and neither does a client logo with no engagement attached to it, named or otherwise. What must be present is a described piece of AI work; what may be absent is the client's name. Every factual claim on a profile must be traceable to public evidence. We describe what is visible and published; we never credit an agency with capabilities it hasn't shown. --- ## The 4 Core Criteria We don't care about flashy awards or big-name logos (unless the engagement behind them is verifiable). We care about one thing: **has this agency shipped AI that works, for real clients?** ### 1. Deployment Evidence (Not Just Demos) Most AI portfolios are demo reels: a chatbot video, a screenshot of a dashboard, a "we integrated GPT" bullet. That tells us nothing. Anyone can wire up a demo in an afternoon. We rank agencies higher when they show the deployed part of the work: - **Production systems:** Is the thing live, with real users or real throughput? - **Stack specificity:** Do they name the models, infrastructure, and integration points, or just say "AI-powered"? - **Measurable outcomes:** What number moved — hours saved, tickets deflected, revenue, error rate? - **Delivery rigor:** Evaluation, monitoring, and what happened after launch. **The reality:** an agency with one plain case study explaining a deployed document-processing pipeline will always outrank an agency with ten polished demo videos and no named client. #### Show, Don't Tell: A Comparison **Agency A (ranked lower)** - **Portfolio:** A show-reel of AI demos and a services page listing every AI buzzword of the season. - **The gap:** No named clients, no deployment detail, no numbers. - **Verdict:** Might be fine — but nothing here is verifiable, so it can't rank on claims alone. **Agency B (ranked higher)** - **Portfolio:** Three case studies with named clients (e.g. "Invoice extraction for a logistics operator — 14,000 documents/month in production"). - **The evidence:** Architecture decisions, evaluation method, what failed in the first iteration, and the measured result. - **Verdict:** Ships real systems. A delivery partner, not a demo factory. We rank Agency B higher every time. This is why our [top-ranked agencies](/agencies) aren't always the ones with the flashiest websites. ### 2. Contextual Expertise A "top" agency isn't top at everything. An agency ranked #1 for **FinTech** in Berlin might be #15 for **Healthcare** — a regulated credit-scoring model and a clinical-documentation tool are different problems with different failure modes. We check portfolios for industry-specific evidence: if they claim healthcare, we expect deployed healthcare work, not a landing page. ### 3. Business Legitimacy The AI boom is full of "agencies" that are one person with a template site. Nothing wrong with freelancers — but you shouldn't pay agency rates for one. We verify: - **Legal entity:** A registered company. - **Team reality:** The people on the About page verifiably work there (we check LinkedIn). - **Physical presence:** A verifiable European address. ### 4. Profile Effort It sounds minor, but it's a real signal. Agencies that complete their profile — exact services, team size, rate range, case studies — tend to be the ones with their operations in order. Claimed and verified profiles demonstrate engagement, which factors into the evaluation. --- ## What Money Buys Here — and What It Never Buys One thing can be paid for on AIAgencies.eu: the Pro membership. It is stated in full in the [terms](/terms); the short version: ### Pro Membership (€49/mo founding rate, €99/mo regular) Listing is free, and the free submission queue is honest but slow: 60-90 days, reviewed in order. Pro includes an **expedited verified review**: subscribing at submission starts a 5-business-day review with evidence verification and badge setup. **The subscription pays for priority, labor, and presence — never the outcome.** Acceptance stays editorial. Rejection is possible; if a listing is rejected, the subscription ends and the first month is retained as the review fee — the review work happened. If we miss the 5-business-day deadline, that fee is refunded automatically: non-refundability never covers our own failure to deliver. For accepted listings, Pro adds instant full-detail lead relay, detailed visit analytics, priority placement in featured slots, an editorial response right, up to six case-study cards on the profile (image, outcome line, summary and link, each reviewed before it goes live), and AI visibility in depth: where AI assistants look in your markets, every answer, the trend across sweeps, and what they say about your firm by name. **It does not buy a higher ranking.** A free-tier agency with stronger deployed work will always outrank a Pro agency with weaker evidence. If we compromised on this, the directory would be useless to you. ### The Link Policy The followed link from a profile to an agency's site is **earned by installing the verification badge** — the mechanic that proves the agency controls its domain. It's available on the free path, and it is never sold as a line item. We don't sell links, we don't sell inclusion, and we don't sell rank. Ever. The same wall stands between our AI-visibility measurement and the register. Probe results, meaning how often AI assistants name an agency when a buyer asks, never affect listing, review or ranking decisions. ### What "Verified" Means The Verified mark on a profile attests two checks and nothing more: the agency confirmed ownership of its listing from a company email address, and the listing passed editorial review before it was published — the same review every listing gets. It is not a business-registration check, a client-reference check, or an endorsement of the work. The evidence for the work lives in the profile itself: the client roster, each entry with its source, and Our Take. Claiming a listing is free, and it never affects rank or score. --- ## Freshness: Every Profile Shows Its Date Stale directories mislead quietly. Ours doesn't: - Every reviewed profile displays its **last review date**, visibly. - Reviews older than six months are flagged internally and re-reviewed on a weekly rota, so the oldest reviews keep moving. - Every acceptance, removal, or approved change republishes the directory — rankings update with the evidence, not on a marketing calendar. - Machine-readable dates (dateModified) are computed from the actual review dates, so what crawlers see matches what you see. ### Editorial Corrections If an agency publishes a significant new case study or we find a verified data error, we update the score immediately. These are corrections based on evidence, not favors. You cannot buy, request, or negotiate a manual adjustment. --- ## Rate Data: Where Every Number Comes From Most agencies never publish an hourly rate on their own site, so most rate figures in any directory are compiled, not quoted. We say which is which: - **Agency-confirmed** — the agency disclosed or confirmed the rate itself, through a submission or a claimed profile. - **Directory-listed** — compiled from the B2B directories the agency lists on, and written only when Clutch publishes the figure (it verifies rates through recorded client interviews) or when two independent directories agree. A lone unverified source is never written into the register, and project budgets published in rate-shaped markup are excluded outright. Non-euro rates are converted at the European Central Bank reference rate on the day of collection. The aggregate statistics — the medians and typical bands on the [rate index](/rates), city pages, and market guides — are computed from these listed ranges and republished with every build. Each cell reports its own sample size and agency-confirmed count, and publishes nothing below an 8-agency floor. --- ## Ranking Tiers and Badges Based on city ranking, agencies are eligible for badge tiers: | Your Rank | Badge | |-----------|-------| | #1 in city | "#1 AI Agency in [City] 2026" | | #2-3 | "Top 3 AI Agency in [City] 2026" | | #4-10 | "Top 10 AI Agency in [City] 2026" | | #11-20 | "Top 20 AI Agency in [City] 2026" | | #21+ | "Featured AI Agency in [City] 2026" | Badges reflect the live ranking and update when it changes. --- ## Why We Don't Publish the Exact Formula We publish what we evaluate, not the mathematical weights. This isn't secrecy; it's anti-gaming. The moment we say "case studies are worth 40 points," agencies start manufacturing case studies to hit the number instead of doing the work. We evaluate the whole picture, the same way every time, and we sign it with a name. --- ## For Agencies: How to Rank Higher 1. **Publish verifiable work.** A named client, the deployed system, the measured result. One deep case study beats twenty thumbnails. 2. **[Claim your profile](/claim)** and complete every field — services, team size, rates, description. 3. **Be specific about your stack.** "AI-powered" is noise; "retrieval pipeline on domain documents with human review" is signal. 4. **Stay current.** Recent deployed work outranks a portfolio frozen in 2023. 5. **Keep your evidence public.** We can only credit what we can verify. Think we got a ranking wrong? [Tell us](/contact). If you can show better deployed work than the agency ranked above you, send the case study. A person reads every email. --- ## Our Commitment We built AIAgencies.eu to help companies find European AI partners with real deployed work — not to sell rankings to the highest bidder. The business model is simple and public: agencies can pay for review speed and for features that help them convert profile visitors into conversations. Inclusion, score, and rank are earned through evidence and are never for sale. If we ever compromise on this, we lose the trust that makes the register worth citing. We won't. --- Source: AIAgencies.eu — the curated register of European AI agencies. Rate provenance: "agency-confirmed" = disclosed via submission or claimed profile; "directory-listed" = Clutch or two agreeing B2B directories (lone unverified sources are never written). Full method: https://www.aiagencies.eu/methodology · Rate index: https://www.aiagencies.eu/rates