We asked AI 54 times about stand building in Spain: 19 of the 20 association members never came up
Company size does not predict whether AI names you. Being readable does.
Across 54 Perplexity answers to real buying questions about trade fair stand building in Spain, 24 of the 40 companies tracked (60%) never appear once. Among them, 19 of the 20 members of the sector association. The one member that does appear shows up in 87% of answers and leads the ranking, without a single block of structured data on its website.
- Sample
- 54 Perplexity answers to 18 buying questions · 40 sector companies tracked across them · 63 websites analysed over HTTP
- Period
- 7 September 2026
- Method
- Perplexity via API, three repetitions per question. Each company is matched by name, domain label and full domain, with word boundaries, and every hit is verified against its literal sentence.
- Language
- Spanish (Spain)
Who shows up when you ask AI for a stand builder
Sixteen of the forty companies tracked appear at least once. The other twenty-four, never.
The top eight, with the share of the 54 answers in which they are named:
| Company | Appears in | Structured data blocks |
|---|---|---|
| Servis Complet | 87.0 % (47/54) | 0 |
| Grupo Diezcarpintero | 57.4 % (31/54) | 1 |
| Standecor | 57.4 % (31/54) | 2 |
| Expomedia | 46.3 % (25/54) | 2 |
| Vissual Events | 44.4 % (24/54) | 1 |
| Adam Expo Stand | 35.2 % (19/54) | 8 |
| Solucionic | 31.5 % (17/54) | 1 |
| Dylunio | 29.6 % (16/54) | 3 |
We do not publish the list of those scoring zero. A low share across 54 answers is not a verdict on a company, and naming anyone in public is not what this exists for.
The company at the top has no structured data
This is the result we did not expect. The company appearing in 87% of answers has zero JSON-LD blocks. The sixth, at 35%, has eight. Among those that appear the average is 1.69 blocks, and among those that do not, 0.96: there is a difference, but it is small, the sample is 40 companies, and the top spot contradicts it outright.
What the leader does have is a concrete, checkable figure in its page title: "Stands para ferias y congresos | 60 años de experiencia" (60 years of experience). And this is what Perplexity answered, verbatim:
"it offers a clear advantage if you are looking for proven experience, since it claims to have 60 years"
It copied the figure straight from the title. The same pattern shows in three others of the top eight ("over 25 years", "over 20 years", a quoted line from their own value proposition), but only one of the sixteen companies that appear carries a figure like that in its title, so this is an observation about one case, not a rule that generalises. We report it because it is what the sentences show, not because the sample proves it.
Three sites ban AI with the exact same file
Of the 63 sites analysed, four block AI crawlers in their robots.txt. Three of them block exactly the same six (GPTBot, OAI-SearchBot, PerplexityBot, ClaudeBot, Google-Extended and CCBot) and the block is identical across all three.
Three independent organisations do not write the same list on their own. It is the block a CDN provider enables by default. Almost certainly nobody decided it, and the effect is that ChatGPT and Perplexity cannot read those sites. Another three return 403 to any automated client requesting their homepage, and AI crawlers are automated clients.
Altogether: 35% of the sector's websites have no structured data at all and only 14% have an llms.txt.
Three in ten citations go to a directory, not a company website
The 54 answers cite 1,024 sources across 54 distinct domains. 70% point to one of the measured companies' own sites; the remaining 30% go to third parties, and a short list dominates there:
- empresite.eleconomista.es — 68 citations
- europages.es — 64 citations
- eventoplus.com, infoisinfo.es and busco1stand.com — the rest of the block
Two general business directories receive more citations than most corporate sites in the sector. For a company that does not appear, that is the cheapest and most verifiable action available: a complete listing on the directory the engine is already reading.
What a company in this sector should take away
- Check your
robots.txtbefore anything else. If it blocks GPTBot or PerplexityBot, nothing else you do counts. It takes two minutes to fix, and you probably did not put it there. - Put a checkable figure where AI reads it. Years, fairs per year, square metres built. The engine reproduces the concrete; it cannot cite the generic.
- Structured data helps, but it is not the ranking. Whoever leads this measurement has none.
- Look at where the engine cites you, not just whether it does. If a directory gets 68 citations in your niche, your listing there is worth more than another page on your own site.
What this finding does NOT prove
- It does not measure the size or the quality of any company. It measures whether an engine names it when answering eighteen specific questions, which is a different and much narrower thing.
- These are 54 answers from a single engine on a single day. Google AI Mode, ChatGPT and Copilot distribute their citations differently, and a measurement next week would give somewhat different figures.
- It does not prove structured data is useless. It proves that in this sample it does not explain who comes first, which is considerably less than it may seem.
- It establishes no causation. That the cited companies appear in certain directories does not prove that being in the directory is what makes them appear.
- The split between association members and outside companies depends on who we put on the list. The twenty members come from the association public directory; the other twenty came out of the answers themselves, so by construction they were more likely to appear.
Frequently asked questions
Why does a large company in the sector not appear in ChatGPT or Perplexity?
Because the engine cannot read it, or because it finds nothing concrete on the site to cite. In this measurement the three most common causes were: a robots.txt blocking AI crawlers, a server returning 403 to any automated client, and a site describing its services in general terms without a single checkable figure.
Does schema or structured data help you appear in AI answers?
It helps the machine understand who you are, but it does not decide the ranking. In this measurement the company appearing in 87% of answers has no JSON-LD block at all, and the one with eight appears in 35%.
How do I know if my robots.txt is blocking AI?
Open yourdomain.com/robots.txt and look for GPTBot, PerplexityBot, ClaudeBot, Google-Extended, OAI-SearchBot and CCBot. If any has a Disallow, that engine cannot read your site. In three of the four blocked sites in this sample the block was identical: the CDN had put it there by default, not the company.
How many times do you need to ask for the measurement to count?
A single pass does not count, because the same engine changes its answer between runs. Here each of the 18 questions was asked three times, hence the 54 answers, and the percentage is calculated over that total.