Kiwop Labs · Evidence-based ranking

AI agencies in Spain: what the assistants recommend

This is not our opinion and nobody paid to be here: it is the count of 200 answers from Claude, ChatGPT (API), Gemini, Perplexity, ChatGPT (chatgpt.com) to real buying questions, in September 2026. Every answer is published in full so anyone can check the count. It updates on its own every month.

How it is built

1. Every month we put 40 questions in Spanish to Claude, ChatGPT (API), Gemini, Perplexity, ChatGPT (chatgpt.com), once each, in a fresh session and with no prior instructions: 15 on AI agencies (buying agents, RAG, LLM integration, audits, the AI Act), 15 from the marketingdirecto.com GEO agency ranking reproduced word for word, and 10 with the short questions people really ask assistants.
2. From each answer we extract the agencies it recommends or lists, in the order they appear, as structured output that can be checked against the full text.
3. The ranking sorts by the number of answers each agency appears in. Ties go to average position within the answer. Size, revenue and quality of work count for nothing here: it measures visibility in the assistants, and nothing else.

Kiwop, which publishes this ranking, appears in 3 answers in the AI block, 1 in the GEO block and 4 in the block of real questions. We publish it anyway: a ranking where the author puts itself first is no use to anyone.

How to read it

• Assistants do not assess agencies: they read what is published and summarise it. An agency climbs here when it appears in the rankings and directories they cite, not when it does better work.
• The big consultancies (Hiberus, Accenture, Minsait, Capgemini) dominate the AI questions because they have more public corpus; the boutiques show up when somebody has listed them.
• Each assistant has its own sources. That is why we publish which domains each one cites: it is the actionable part of the data.

Where the answers come from

Domains most cited by the assistants across this set of questions. If you want to appear in the ranking, this is the list of sites you need to be on.

What this ranking is not

• It is not a quality ranking: we have not audited any agency's work.
• It is not a census: one pass a month with one answer per question. It is a sample, and it should be read as one.
• It is not each provider's exact consumer app: four assistants are measured through their API with web search, and the fifth is ChatGPT on chatgpt.com with no session. Model and date go with every data point.
• Nobody pays to appear and nobody can ask to be removed. If you think there is an extraction error, the full answer is published: write to us with the link.

Updates

It is recalculated on the 1st of every month from the Kiwop Labs GEO baseline. The historical series, prompt by prompt, is on the baseline page.

Data

• GEO baseline: method, full answers and monthly evolution
• This month's dataset (JSON) · 2026-09.json
• Scripts to reproduce it (GitHub)

How to cite

Kiwop Labs (September 2026). AI agencies in Spain: what the assistants recommend. https://www.kiwop.com/en/labs/ai-agencies-spain-ranking

Data and text under a CC BY 4.0 licence: you can use and quote them, with a link to this page.

← Kiwop Labs