ProtForge

Updated: 17 Nov 2026 · next edition: December 2026

AI visibility index of Spanish SMEs, November 2026

We asked ChatGPT, Claude and Perplexity «what is the best X in Y» for 29 Spanish SMEs across six sectors and counted how often each one is named. The sample is small and not random, and we state its limits. The data is ours and it is open.

See the findings Download the CSV

Six findings

All figures are for Spanish SMEs. The sample size of each is in brackets.

26.9%of Claude answers name the business (39 of 145)

More than half show up sometimes, nearly half never

Of the 29 SMEs, 15 were named at least once across five questions and 14 never appeared. Only 3.4% of Claude answers (5 of 145) linked the business's own website as a source.

0 of 7dental clinics named by Claude (0 of 35 answers)

Sector matters more than the website

Dental clinics never appeared in Claude, while marketing agencies (3 of 3) and other health and aesthetics clinics (5 of 7) came up often. With 3 to 7 businesses per sector this is a hint, not a conclusion.

22.2%ChatGPT web (12 of 54); Perplexity web 10.9% (6 of 55)

Each engine names different businesses

Of 11 businesses Claude had not named, ChatGPT named 4 and Perplexity named 2. Visibility cannot be measured on a single engine. This group was chosen precisely because it did not appear in Claude, so its ChatGPT and Perplexity figures are probably lower than a normal sample would give.

72average technical score out of 100 (median 72; range 51 to 93)

The technical side is fine except structured data

By area, the average is 88 for trust signals, 82 for readability without JavaScript, 75 for on-page content, 71 for crawlers, 62 for discoverability and 55 for structured data, the lowest. None of the 29 sites blocks AI crawlers in robots.txt, so access is not the problem.

41%have an llms.txt (12 of 29)

llms.txt: nearly half already have one

Businesses with an llms.txt were named by Claude in 40.0% of answers and those without in 17.6%. It is an association, not a cause: the sectors that have it most (marketing, tech) live off online marketing and probably stand out for other reasons too.

28%declare a local-business type (8 of 29) and 28% have FAQPage (8 of 29)

Nearly all have JSON-LD, few the right kind

28 of 29 sites publish some JSON-LD, but in many cases only the generic markup from the CMS (WebSite, WebPage). Only 8 declare a local-business type (LocalBusiness, Dentist, MedicalClinic, LegalService and similar) and 19 an Organization.

By sector

Claude sample, 5 questions per business. With 3 to 7 businesses per row, do not compare rows by decimals: they only show the order of magnitude.

SectorBusinessesClaude mentionsAvg. technical scoreWith llms.txtWith FAQPage
Dental clinics70%6929%43%
Other clinics (aesthetics, fertility, physiotherapy)749%7129%29%
Law firms532%7320%0%
Education and languages45%6625%0%
Marketing agencies360%76100%33%
Other (real estate, industrial, software)327%86100%67%
All Spanish SMEs2927%7241%28%

Mentions: share of the 5 answers per business in which Claude names the business (145 answers in total). Rows are ordered by sector, not by result.

Three engines, subsample of 11 businesses

Engine (web version, logged out)AnswersMention rateCitation rateBusinesses named
ChatGPT5422.2%13.0%4 of 11
Perplexity5510.9%5.5%2 of 11
Claude550%0%0 of 11

This subsample (6 dental clinics, 3 education centres, 1 aesthetics clinic and 1 law firm) was re-audited precisely because Claude had not named them, so its 0% is by construction and not comparable with the table above. One of the 55 ChatGPT answers failed and is not counted.

What to do

Five things, from cheapest to slowest. None guarantees a mention, but all of them move signals you control.

  1. Publish an llms.txtA short file at the site root saying who you are, what you sell, where, with links to your key pages. An afternoon's work.
  2. Declare your business with structured dataJSON-LD with the right type (LocalBusiness or a subtype such as Dentist, LegalService or MedicalClinic), address, opening hours and FAQPage on service pages. 72% of the sample does not declare a local-business type.
  3. Earn third-party mentionsOnly 3.4% of answers link the business's own site: in the answers we reviewed, many sources are directories and comparison pages. Get listed in the directories of your sector and city.
  4. Look after reviewsRecent, answered reviews on Google and sector directories, with the service name and the city in the text.
  5. Put prices on your siteBuyer questions include «on a tight budget»: a page with explicit prices or ranges and an update date gives engines something concrete to cite.

Method and limits

Sample. 33 businesses audited on 16 and 17 November 2026, of which 29 are Spanish SMEs (12 in Valencia, 8 in Barcelona, 6 in Madrid and 3 in other cities) and 4 are US businesses that are left out of every figure on this page. We picked them as prospective clients for our audit: the sample is not random and over-represents SMEs with their own website and a local presence in large cities.

Prompts. For each business, five buyer questions that do not name the brand, built from its category and city («What are the best companies for…?», «I need help with… Who do you recommend?», «How do I choose a provider for…?», among others). One of the five asks for alternatives to a known competitor, which can push the engine to list competitors. The prompts are in Spanish.

Engines. Claude for all 29 businesses (145 answers). For 11 of them, also ChatGPT and Perplexity in their web versions, logged out, so with no history or personalisation. Google AI Overviews was left out because it blocks automation. These engines are not deterministic: the same question can return different answers on the same day.

What we count. Mention: the business is named in the answer. Citation: its domain appears among the sources. Name extraction is heuristic and only partly reviewed by hand.

Technical data. One crawl per business of up to 25 pages: robots.txt, llms.txt, JSON-LD types and a 0-100 score across six areas, with aieo-audit, the tool we use in our audits. JSON-LD counts if it appears on any crawled page.

Limits.

The aggregate data is in indice-visibilidad-ia-pymes-2026-11.csv. It contains no names of businesses or of competitors that appear in the answers.

FAQ

What share of Spanish SMEs show up in ChatGPT, Claude and Perplexity answers?

In our sample of 29 Spanish SMEs, Claude named the business in 26.9% of answers, and 14 of the 29 never appeared in any of the five questions. In a subsample of 11 businesses Claude had not named, ChatGPT named them in 22.2% of answers and Perplexity in 10.9%. The sample is small and not representative, so the figures describe this sample, not Spanish SMEs in general.

Does having an llms.txt make AI engines recommend you more?

In our data, businesses with an llms.txt (12 of 29) were named by Claude in 40.0% of answers versus 17.6% for those without one. That is an association, not proof of cause: the sectors with the most llms.txt files (marketing agencies and tech companies) live off online marketing and probably stand out for other reasons too. An llms.txt is cheap to add and worth having, but it does not replace third-party mentions.

How was AI visibility measured?

For each business we sent five buyer questions that do not name the brand (for example, «What are the best companies for a trustworthy dentist in central Valencia?», in Spanish). A mention means the business is named in the answer; a citation means its domain appears among the sources. We used Claude for all 29 businesses and, for 11 of them, also ChatGPT and Perplexity in their logged-out web versions, on 16 and 17 November 2026.

Can I download the data?

Yes. The CSV with aggregate figures by sector and by engine is at /data/indice-visibilidad-ia-pymes-2026-11.csv. It contains no company names. The figures are generated by aieo-audit, the tool we use in our audits.

What about your business?

The AI visibility audit measures the same thing as this index, but with 50 to 100 questions about your sector and your competitors on every engine, plus the technical review and a prioritised plan. $1,500 (1.200 € + VAT for Spanish clients), report and a 45-minute call within 5 working days.

See the AI visibility audit ($1,500)