Back in June 2025, we promised you an article about measuring your LLMS.txt file's effectiveness. It took us 14 months to write it, and honestly, that delay was the most educational part.
We waited because the tools to answer the question didn't exist yet. Most of them arrived this spring. And while we waited, we accidentally ran the perfect experiment on our own website.
Let me show you what we found, what you can actually measure in 2026, and the part nobody can measure no matter what their dashboard says.
The experiment we didn't mean to run
From late February until August 19 of this year, our own llms-full.txt file told every AI system that read it about three service plans we had retired, at prices we no longer charge. Our llms.txt had 4 broken links in it since March.
The total business consequences of publishing wrong prices in a machine-readable file for almost half a year: zero. No confused callers. No emails asking about the old plans. Nothing.
That's the honest baseline for this whole topic. If AI systems were reading these files and repeating them to customers, we'd have heard about it. The silence was the measurement.
So when we say the tools below are worth 15 minutes a quarter, we mean exactly that. Fifteen minutes. Not a dashboard you live in.
What you can measure now
Three real tools arrived in 2026. Each one shows you a different slice, and each one has a blind spot you should know about before you trust it.
1. Search Console's AI performance report
On June 3, 2026, Google launched a dedicated AI performance report in Search Console. It shows how often your pages appeared in AI Overviews, AI Mode, and generative features in Discover, broken down by URL, country, and device.
What it won't show you: clicks, the prompts people typed, or your position. Impressions only. And remember from the first post in this series: Google ignores your llms.txt file entirely, so this report measures your regular content's AI visibility. Your llms.txt has no influence on it, by Google's own written statement.
It rolled out starting in the UK, so if you don't see it under Performance yet, it's coming.
2. GA4's AI Assistant channel
On May 13, 2026, Google Analytics added a native AI Assistant channel that automatically groups visits from ChatGPT, Gemini, and Claude.
Now the blind spots, and they're big. Perplexity still lands in Referral. AI Overview clicks count as Organic Search. And here's the one that matters most: somewhere between 35% and 70% of visits from AI tools arrive with no referrer at all, which means GA4 files them under Direct. Whatever AI traffic your GA4 shows, the real number is probably about double.
3. Your server logs (or Cloudflare)
This is the only place you can see whether anything fetches your llms.txt file, because crawlers that read a text file never trigger analytics. If your site sits behind Cloudflare like ours does, its bot analytics will show you which AI crawlers hit your site without touching a log file.
Set your expectations before you look. Ahrefs studied 137,210 domains and found 97% of llms.txt files received zero requests in a month. If your log shows crickets, you're in the 97%, and that's the normal reading in 2026.
And if you do see a fetch, hold the champagne. A bot requesting your file proves the bot requested your file. The cats.txt experiment we covered in post 1 showed that AI crawlers will happily fetch, and Google will happily index, a file about fictional office cats.
What no tool can measure
Picture the actual win everyone's chasing. Someone asks ChatGPT for a WordPress maintenance company they can trust. ChatGPT mentions you. The person picks up the phone and calls.
That sequence is invisible in Search Console, invisible in GA4, invisible in your server logs. No click happened. No referrer exists. The most watched dashboard in the world would show you nothing.
This is why the plumber who gets recommended by an AI assistant and books the job will never appear in any AI traffic report, and why anyone selling you “complete AI visibility analytics” deserves a skeptical squint.
The fingerprint trick that actually works
You can't track the phantom phone call with software, so you track it the old-fashioned way: leave fingerprints that could only have come from one place.
Yoast has run this play since 2025. Their llms.txt file contains a discount code, LLMTXT15, that exists nowhere else. Every redemption is a provable AI-file conversion. The code is still in their file as of this week, which tells you the experiment costs them almost nothing to keep running.
Your version of the same play:
- Put a unique offer code in your llms.txt file that appears nowhere else on your site
- Give the links inside the file their own UTM parameters so any click-through is labeled
- Keep asking “how did you hear about us?” on your intake form, and add “an AI tool (ChatGPT, Claude, etc.)” as an option
That last one is doing the heaviest lifting in 2026. A one-line form field catches the phone-call conversions that a billion dollars of analytics infrastructure cannot see.
Why bother measuring traffic this small?
Because the little there is converts like nothing else you have.
Ahrefs measured AI-referred visitors at 0.5% of their traffic and 12.1% of their signups. Someone who arrives after an AI tool explained exactly why you fit their problem shows up pre-sold. Small stream, absurdly warm.
That's the real reason to spend the quarterly 15 minutes. You're watching a small channel that punches far above its size, waiting to see when it stops being small.
Your quarterly 15-minute measurement routine
- Open Search Console's AI performance report and note the impression trend. Up, down, flat. That's it.
- In GA4, check the AI Assistant channel and skim Referral for Perplexity. Mentally double whatever you find.
- Load yourdomain.com/llms.txt in a browser and confirm it still loads and still tells the truth. (Ours didn't for five months. Learn from us.)
- Count redemptions of your fingerprint code and check your intake form's “how did you hear about us” answers.
- Write the four numbers down somewhere. The trend across quarters is the actual signal.
Your measurement questions answered
Does Search Console show clicks from AI Overviews?
No. The AI performance report shows impressions only: no clicks, no prompts, no positions. Click data on AI-referred visits lives in GA4, and only for the visits that carry a referrer.
Why does GA4 show so little AI traffic?
Partly because AI traffic is genuinely small for most sites, and partly because GA4 can't see a lot of what exists. Up to 70% of AI-tool visits arrive without a referrer and get filed under Direct, Perplexity gets filed under Referral, and AI Overview clicks blend into Organic Search. Treat the AI Assistant channel as a floor, never a total.
Can I see whether ChatGPT read my llms.txt file?
Only in your server logs or a Cloudflare-style bot dashboard, and the overwhelmingly likely answer is that it hasn't. 97% of these files get zero requests in a given month. Also remember that a fetch proves interest from a crawler, and nothing more.
Is AI traffic even worth tracking at this size?
Yes, because of the conversion quality. Ahrefs' own numbers: 0.5% of visitors, 12.1% of signups. The job is watching a warm channel closely enough to catch the moment it grows.
What's the one thing to do if I only do one?
Add “an AI tool” to your intake form's how-did-you-hear question. It's free, it takes 2 minutes, and it's the only method on this page that catches the conversions no analytics tool can see.

You didn't start your business to become a WordPress technician, yet here you are treating your most critical business asset like a hobby project. While you're spending hours on updates and crossing your fingers that nothing breaks, your competitors are focused on what actually makes money.
👋 I'm Jeane Sumner, and Website HQ doesn't just host your site – we armor it, optimize it, and monitor it so you can get back to running your business instead of fixing it. Book your Site Care strategy call and never worry about your website again.
In this series:
- Why Your Website Needs an LLMS.txt File (And What the Heck That Even Means)
- What Industry Leaders Put in Their LLMS.txt Files (And How to Steal Their Strategies)
- How to Create Your First LLMS.txt File: A Step-by-Step Guide
- The LLMS.txt Domain Dilemma: Where Should You Actually Put Your File?
- How to Measure AI Traffic (And Whether Your LLMS.txt File Is Doing Anything at All) –This article!
Google now lets you pick which sites get priority in your Search and AI results. If this series helped, add us. Two clicks, and you can undo it anytime.
