The llms.txt question refuses to die. I got asked twice last week, once by a founder who'd been quoted $1,200 for an "llms.txt optimization package," and once by an engineer who'd read that GPT-5.6 changed everything and wondered if the file finally mattered now. Same file, opposite hopes. So let's settle it properly, with the evidence we have, including our own.
Short version, for the people who'll skim: no, llms.txt is not worth strategy or money after ChatGPT 5.6, and 5.6 actually made the case weaker. Twenty minutes to create one as a courtesy file? Sure, knock yourself out. Anything beyond that is spending real hours on a file with no measured effect. Now the long version, because the reasoning is more useful than the verdict.
What 5.6 Actually Changed About Source Selection
Here's the thing people get backwards. When GPT-5.6 landed in August and citation patterns went sideways, a chunk of the SEO world concluded that ChatGPT must be reading new signals, so maybe now the courtesy file counts. The tracking data says the opposite happened.
Post-5.6 retrieval got more conservative, not more exotic. Site-scoped fan-out queries jumped from under half a percent to 17-23% of searches: the model increasingly decides which trusted domains to search before deciding what to read. "Official" became one of its favorite terms to append to its own retrieval queries. Cited domains consolidated even as retrieved URLs doubled. The whole mechanism shift points one direction: indexed pages on verifiable domains, weighted by trust.
Now ask where llms.txt sits in that machinery. It's not in retrieval indexes; the engines index your pages, not your courtesy files. It's not a trust signal; anyone can write one, which is precisely why it can't carry trust. And it's not what the fan-out queries hit. The file was designed as a friendly map for AI systems that choose to fetch it, and the systems deciding your citations in 2026 assemble answers from a different room entirely.
What llms.txt Was Actually Designed For
Quick detour into the origin story, because half the confusion is people arguing with a file that never claimed what they're mad it doesn't do.
The llms.txt proposal came out of the developer-tools world in 2024 as a courtesy map: a markdown index at your site root so an AI system with a limited context window could find your important pages without crawling everything. Think of it as a README for robots. For documentation sites, where an agent genuinely benefits from "here are the twelve pages that matter, skip the changelog," that's a sensible, modest idea, and it still is.
What happened next is pure SEO folklore mechanics. The file was easy to create, visible to competitors, and adjacent to a scary trend, which is the exact recipe for a tactic going viral regardless of effect. Within months, "add llms.txt" appeared on every AI-SEO checklist as though it were a ranking submission form, agencies added it to audits, and the proposal's actual, humble scope got buried under promises its own authors never made. The file didn't fail at its job. It got assigned a different job by people selling checklists, and it fails at that one.
Which means the honest question was never "does llms.txt work?" It's "does a courtesy index for context-limited agents affect citation selection in retrieval-augmented answer engines?" Phrased accurately, the answer stops being surprising.
The Evidence Pile, Honestly Stacked
Let's be real about how much evidence there is, on both sides.
Against: we ran llms.txt across 12 sites for 90 days and measured no citation lift, on any engine, at all. Google has flatly said it doesn't use the file. And the adoption studies keep finding the same gap: sites add the file, crawler logs show the big engines mostly not fetching it, and citation behavior doesn't move. That's a first-party test, the largest engine's own statement, and log data, all agreeing.
For: it's nearly free, and a long tail of smaller agents and tools genuinely does fetch it when present. Some developer-facing crawlers read it. Someone's homegrown research agent might use yours this afternoon. That's real, and it's also the entire case.
Funny enough, the strongest thing I can say for llms.txt is that it's a lottery ticket that costs twenty minutes. The strongest thing I can say against it is that an industry sprang up selling lottery-ticket optimization, and that industry is a preview of GEO spam generally: whenever a tactic is easy to deliver and impossible to measure, someone will invoice for it.
200+ SaaS teams already track their AI citations.
They know exactly when ChatGPT mentions their brand, and when it stops. Do you?
So: Worth It?
My actual answer, in three tiers.
Creating one, once: fine. Twenty minutes, a tidy little index of what your product is and where the important pages live. It might help some agent someday. It will not move your citations, and you should not expect it to. File it mentally next to a favicon: correct to have, silly to strategize.
Maintaining one as a priority: no. If updating llms.txt appears on anyone's quarterly goals, something has gone wrong with the quarter. The file describes your site to systems that mostly don't read it, while the systems that decide your visibility read your actual pages, which brings us to the opportunity cost.
Paying for llms.txt optimization: absolutely not. A $1,200 package for a file with no measured effect isn't optimization, it's a tax on anxiety. And I'd go one further: a vendor who leads with llms.txt just told you how they choose deliverables, and it isn't by effect size.
Spend The Afternoon Where The Engines Actually Read
The fair question back is "fine, then what do I do with the twenty minutes and the anxiety?" Three trades, each with a measurable mechanism behind it.
Rewrite your openings. Most ChatGPT usage answers from your title and roughly the first 200 characters of body text, without opening the page. An opening that states the answer plainly is worth more than any courtesy file could theoretically be, because it's literally the text the answer gets built from. Twenty pages, one afternoon, done.
Verify access instead of describing it. An llms.txt that says "welcome, AI systems" on a site whose CDN bot protection is challenging every AI fetcher is a doormat in front of a locked door. Fetch your key pages as the AI user agents and confirm content comes back. This one check finds more lost visibility than every courtesy file ever written.
Fix what the trust-seeking retrieval actually checks. Consistent entity facts across your site and profiles, named authors, parseable pricing. Post-5.6 machinery verifies brands against corroboration, and that's the layer where an afternoon compounds.
For what it's worth, we watch this from the measurement side every week, and I have yet to see a citation shift that traced to a courtesy file. I've seen plenty trace to an unblocked crawler, a rewritten opening, or a review profile that finally agreed with the homepage. The boring stuff keeps winning, which is the most 2026 sentence in this whole post.
Answering The Boss Who Read About It
Between you and me, most people asking about llms.txt aren't really asking about the file. They're asking one of three other questions wearing its clothes.
Sometimes it's cheap-win anxiety: "is there a twenty-minute thing we're missing while competitors get cited?" Fair worry, wrong file; the twenty-minute thing you're missing is almost certainly the opening-paragraph rewrite or the CDN check above. Sometimes it's audit theater: an agency deliverable listed the file as missing, in red, and now it needs an answer. And sometimes it's just the boss forwarding a LinkedIn post with "should we be doing this?"
For that last one, here's the reply that works, copy-paste free of charge: "We evaluated it. The largest engine says it doesn't use the file, a 12-site 90-day test measured no effect, and the retrieval systems that decide our citations read indexed pages, not courtesy files. We added one anyway since it costs nothing, and we spent the real effort on openings, crawler access, and entity consistency, which are the levers with measured effects." That's four sentences, it's all true, and it converts the anxiety into the work that pays.
How We'd Know If This Verdict Expires
One more thing, because "no" verdicts in this space have a shelf life and I'd rather state the expiry test than pretend to permanence. The verdict flips if any major engine documents the file in its retrieval pipeline, or if fetch logs at scale start showing the big crawlers reading it and citation changes correlating with its contents. Either would be visible within weeks in crawler logs and tracking data, and neither has happened through two model generations and a retrieval rewrite.
So we keep re-checking, the same way everything in AI search gets re-checked: on a schedule, against data, with the weekly per-engine tracking doing the watching. If the file ever starts mattering, you'll read the update here. I wouldn't hold the afternoon open for it.
How often does ChatGPT mention your brand?
Most founders have no idea. The answer might surprise you.
The One-Line Answers
Is llms.txt worth it after ChatGPT 5.6? As a twenty-minute courtesy: harmless. As a strategy: no. As a paid service: run. And if the question on your mind is really "how do I know what's working," that's a tracking question, and no static file was ever going to answer it.
15 hours a week manually. Or 15 minutes with RankControl.
Track citations, monitor competitors, and fix content gaps across every AI search engine. Automatically.
