Somewhere in the last year, "add an llms.txt file" became the AI-era equivalent of "just submit your sitemap to Google". A quick, discrete, technical fix that supposedly earns a business a place in ChatGPT's answers. It's easy to sell, easy to implement, and easy to point to as proof you're "doing AI SEO".
We had the data to actually check whether it works. For our AI Visibility Report, we audited 10,000 US local businesses through the Insites 360-degree platform and correlated a full set of "AI readiness" signals, llms.txt, structured data, FAQ schema, against real ChatGPT and Perplexity outcomes. The honest answer: these tactics help a little, at best, and several of them don't move the needle at all.
What the data actually shows
Start with llms.txt itself. Businesses found by both ChatGPT and Perplexity had it in place 29.2% of the time, versus 24.7% for businesses not found by either. That's a real gap, but a small one, and it means plenty of businesses that took the time to add an llms.txt file still weren't surfaced by AI.
The pattern repeats across the category:
| Factor | Found by both platforms | Not found | Correlation |
|---|---|---|---|
| LLMs.txt present | 29.2% | 24.7% | Weak |
| FAQ structured data present | 9.4% | 5.6% | Moderate |
| Local structured data present | 36.8% | 28.1% | Moderate |
| Structured data present (any) | 73.1% | 64% | Weak |
| Website marked "AI-optimised" | 6% | 2.3% | Moderate |
When we looked at recommendation specifically, rather than just awareness, the picture got even weaker. FAQ structured data and "AI-optimised" website status showed no meaningful relationship at all with whether a business was actually recommended in unbranded queries. Adding schema didn't stop a business from being invisible when it mattered most. It's the same story we found when we tested website speed in isolation: another technical checklist item that barely moves the outcome.
For contrast, here's what did correlate strongly across the same dataset: Google review volume (133.4 average reviews for businesses found by both platforms, versus 10.7 for neither), local pack presence (43.4% versus 10.1%), and directory listing coverage. None of those are AI-specific tactics. They're the same fundamentals that have driven local SEO for years.
Calling it what it is
We said this plainly in the report itself: "We're already seeing the early signs of 'snake oil' creeping in. Quick fixes, AI hacks, silver bullets. The reality is, those don't work here. AI visibility isn't something you can game with a trick or a one-off optimisation. It's built on fundamentals, presence, consistency, relevance and reputation, and those take ongoing effort."
That's worth sitting with, because it cuts against a genuinely tempting sales pitch. "Add this file and get found by ChatGPT" is a much easier thing to sell than "build up your review base, keep your listings consistent, and maintain a content-rich website over the next six months". But only one of those is what the data supports.
So is llms.txt worth doing at all?
Yes, just not for the reason most people are selling it. Adding structured data, FAQ schema or an llms.txt file costs very little and shows up as a small positive correlation across the board. Treat it the way you'd treat a sitemap submission: a sensible, low-effort piece of housekeeping, not a strategy.
What it can't do is substitute for the fundamentals. A perfectly marked-up website with ten reviews and inconsistent directory listings will still lose out to a messier site backed by a strong Google Business Profile and a real review history.
What to prioritise instead
If a client asks what to do about AI visibility, the order of operations the data supports is:
- Google Business Profile first. Reviews, completeness, photos, claimed status, this is the strongest correlating category in the entire dataset
- Local presence second. Local pack visibility, directory coverage, and consistency of NAP data across listings
- Content depth third. A larger, more frequently updated website with genuine service and location content
- Technical AI readiness last. Structured data, FAQ schema and llms.txt as a low-cost bonus, once everything above is in place
Insites' AI visibility audits are already weighted in this order, so agencies get a prioritised action list instead of a flat checklist of forty equally-weighted technical items.
AI-specific optimisation isn't worthless, it's just not where the leverage is. The businesses actually winning in ChatGPT and Perplexity are the ones that were already winning at the fundamentals, reviews, consistency and presence, long before "AI SEO" was a line item.
This analysis is drawn from Insites' AI Visibility Report, based on a clean audit of 10,000 US local businesses across ChatGPT and Perplexity. Run a free AI visibility check on any business in seconds.
Andrew is the co-founder and CEO of Insites, the AI visibility platform that agencies, telcos and media companies use to show local businesses how search engines and AI tools like ChatGPT and Gemini find, trust and recommend them. He has spent more than 15 years in digital marketing and UX, and co-hosts The Unusable Podcast on UX design.
