Best agencies for AI document search (RAG) in Spain 2026: expertuno Score ranking

Methodology

What it is. The expertuno Score is a 0–100 score, the same for every profile, that adds up five components with public weights: Google reputation (30 points), declared technology (15), complete profile (15), website and verification (15) and public proof (25). Google reviews account for 30 of the 100 points; the rest depends on the technology each profile declares, how complete and verifiable it is and the public proof of its work. The ranking is sorted by the total; ties go to the higher weighted Google rating and then to the profile with more reviews. Each row shows the total and its breakdown by component.

Google reputation (30 points). It starts from a weighted (Bayesian) average that balances rating and number of reviews: each profile gets 20 “virtual” reviews at the ranking's average rating (C = 4.78 out of 5), so with few reviews its rating stays close to that average and the more reviews it has, the more its own rating counts (with 20 reviews, half; with 180, 90%). Weighted rating = (v ÷ (v + 20)) × R + (20 ÷ (v + 20)) × C, where R is the profile's rating and v its number of reviews. That weighted rating earns up to 25 points on a linear scale from 4.0 (0 points) to 5.0 (25 points), and review volume adds up to 5 more points (5 with 200 reviews or more). Without a Google rating, this component is 0.

Declared technology (15 points). It comes from the technologies and specialties each profile declares: at least one large language model (LLM: OpenAI, Anthropic Claude, Google Gemini, Mistral, Meta Llama, Azure OpenAI…) earns 5 points; at least one automation or orchestration platform (n8n, Make, Zapier, Power Automate, LangChain…), 3; at least one data or cloud platform (AWS, Microsoft Azure, Google Cloud, Snowflake, BigQuery, Databricks, Power BI…), 3; five or more declared technologies, 2; and three or more specialties, 2. Each technology is recognised by its exact name from closed lists that are the same for every profile. It measures what the profile declares; it does not audit how the technology is used in projects.

Complete profile (15 points). Spanish description of at least 300 characters (3 points), tagline (1.5), three or more specialties (2), three or more technologies (2), at least one sector (1.5), team size (1.5), founding year (1.5) and logo (2).

Website and verification (15 points). Website working at the directory’s last check (6 points), profile claimed by its owner (4) and a current identity verification (5).

Public proof (25 points). Case studies published on the profile: one earns 5 points, two earn 8 and three or more earn 12 (case studies are currently published from the profile with the Pro plan). Public evidence on the profile: each case study or publication earns 3 points (up to 9) and each certification, talk or open-source project earns 2 (up to 4). Only evidence marked as public counts.

Upcoming criteria. Under study: domain authority and press mentions; they will be added once we have a verifiable, public source.

Who is included. Published profiles the directory classifies as AI specialists (AI core: 506 of 790 published), excluding demo listings; adjacent providers —marketing, training or IT services that offer some AI— are listed in the directory but not ranked. There is no minimum number of reviews: of the 506 ranked profiles, 304 have Google reviews, 306 earn points for declared technology and 1 has public proof (case studies or evidence). This page shows the 58 that declare the “Document search” specialty.

Sponsorship. Sponsorship does not change any score or position. The “Sponsored” card sits outside the numbering and, if the sponsored profile also appears in the ranking, its row carries a “Sponsored” label. NextAI is a brand of the same publishing group as expertuno; it appears as an identified sponsor and is scored with the same public formula as everyone else.

How to improve a score. Claim your profile and complete it (description, tagline, specialties, sectors, team, founding year and logo); declare the technologies you work with (language models, automation, data and cloud); keep your website working and verify your identity; publish case studies and public evidence (case studies, publications, certifications, talks or open source); and ask your clients to rate you on Google. Scores are recalculated with every data update. Claim my profile ↗Plans for professionals ↗

Sources and date. Rating and number of reviews: each provider's public Google Maps listing (expertuno does not publish reviews of its own). Technologies, profile, website, verification, case studies and evidence: expertuno directory data, as declared or published by each profile. Data as of 2 November 2026.

Limits. The score measures public reputation, declared technology and how complete and verifiable each profile is, not the technical quality of each project. Check each profile’s specialties, evidence and verification before deciding.