llms.txt Is Not a GEO Shortcut. Evidence Pages Still Carry the Weight
Ahrefs found that most llms.txt files in its dataset received no traffic, reinforcing the value of crawlable, source-backed pages.
Author
Tim Walsh
Reading time
7 min read
Updated
2026-07-10
01
The shortcut story is weaker than the data
Ahrefs analyzed 137,000 domains using Ahrefs Web Analytics and reported that 28% published an llms.txt file, while 97% of those files received zero traffic in May 2026.
That does not make llms.txt useless for every future workflow, but it makes one point practical today: strong public pages remain the most reliable asset.
Dataset
137K domains
Ahrefs studied more than 137,000 domains using its web analytics dataset.
Adoption vs usage
28% publish / 97% unread
28% of studied domains published valid llms.txt files, but 97% of those files received zero requests in May 2026.
Geolity angle
Page evidence first
The practical content system keeps crawlable evidence pages ahead of machine-readable shortcut files.
02
The adoption number and the usage number tell different stories
The most useful contrast in the study is the distance between adoption and actual requests. A visible portion of sites added the file, but almost all of those files were not requested during the measured month.
That is exactly the kind of distinction GEO content can make. A tactic can be popular because the market is anxious, while observable AI-search visibility still comes from evidence and retrieval behavior.
For Geolity, the article teaches teams to connect implementation with evidence: what changed in prompts, what source was cited, what page was reused, and what traffic or crawler behavior can be observed?
Source signal matrix
| Public signal | What the source shows | Geolity advantage |
|---|---|---|
| Valid llms.txt adoption | Ahrefs reports that 38,360 of 137,000 studied domains published valid llms.txt files. | Adoption becomes more useful when paired with core pages that can support AI answers. |
| Zero-request files | Ahrefs reports that 97% of valid llms.txt files saw no requests in May 2026. | A public report can treat llms.txt as a support signal while keeping retrieval evidence and page quality as the main story. |
| AI bot request mix | Named AI bot categories made up 19.5% of requests among the small subset of files that received traffic. | Machine-readable files may be useful for agentic workflows, but prompt visibility still depends on pages, citations, and answer evidence. |
03
Google's guidance points the same direction
Google's own generative AI search guidance says llms.txt is not required for Google Search generative AI features and that Google Search ignores it for visibility or ranking purposes.
The shared lesson is simple: build the page that can carry the answer before optimizing the map that points to it.
Geolity product advantage matrix
| Product layer | Geolity advantage | Reader value |
|---|---|---|
| Canonical evidence | Geolity keeps the canonical HTML page at the center of GEO work. | AI retrieval systems get a real page to cite, not only a machine-readable pointer. |
| Crawler signal | Geolity can pair machine-readable assets with observable crawler and referral behavior. | Teams understand which technical assets create visible activity and which ones are support layers. |
| Agent support | Geolity treats llms.txt as optional agent infrastructure beside stronger public pages. | Teams can support emerging agent workflows while keeping measurement tied to real AI visibility. |
04
The files that are fetched are mostly not proof of AI-search demand
Among files that did receive traffic, Ahrefs reports that 96% of requests came from bots and that 77% of bot requests were not from AI tools.
The named AI categories are still interesting: agents and agentic infrastructure represented 10.5% of requests, training crawlers 5.3%, assistants 2.5%, and retrieval bots 1.1%.
This suggests a more nuanced strategy: keep machine-readable assets as future-facing infrastructure, but do not confuse crawler curiosity with buyer-facing AI visibility.
05
What Geolity measures as the stronger signal
A useful GEO review shows whether buyer prompts connect to answerable evidence: clear claims, comparison proof, product context, methodology, and source-backed category pages.
If an llms.txt file exists, Geolity can log it as a supporting asset while keeping prompt outcomes and cited pages as the primary visibility evidence.
The reporting view shows a clear hierarchy: prompt outcome first, cited source second, owned-page readiness third, and optional machine-readable assets last.