Hook: Twelve approved pieces sit in a folder, and none of them is anywhere an AI can fetch. Building the host and the files yourself costs EUR 720 to EUR 2,160 of developer time, plus a CMS ticket every time you publish.
What You'll Be Able To Do
- Put this month's approved content on a subdomain of your own domain in one pass
- Tag every piece by audience and topic, both taken from your scan
- Hand IT two files and one line of script, once, with no ticket per publish after that
- Open the live llms.txt in a browser and send the link to anyone who asks
- Count the people who landed on your pages out of an AI conversation, and the crawlers that pulled them
- Take one piece down with a switch
Total Time: 6:50 | Difficulty: Advanced
Getting the folder live takes a CMS ticket for the website, then a second ticket for the files AI crawlers look for. Two queues, neither of them yours, both re-opening every time you publish. Once the pages are live, nobody can tell you whether a model came to read them: crawler traffic gets filtered out of your analytics by design, and a visitor from a chat app often carries no referrer.
What it costs today
A developer at EUR 90 an hour.
| Line | Arithmetic | Cost |
|---|---|---|
| Crawlable content host plus correct llms.txt and robots.txt, one time | 8 to 24 h × EUR 90 | EUR 720 to EUR 2,160 |
| Same work bought in as a technical GEO sprint | agency quote | EUR 5,000 to EUR 15,000 |
| CMS ticket per publish, from then on | recurring, rarely budgeted | |
| Rankfor Marketer | EUR 196 per month |
Total: EUR 720 to EUR 2,160 of developer time for the setup, against EUR 196 of Marketer.
What the video shows
Ten shots, in order, so a reader who cannot watch still learns the flow.
- Publishing opens on three routes, Manual, AI Content Layer and Paid Boost, with this month's box and a green "Loop active" pill.
- Wizard step 1: filter the assets by source, toggle on what is approved, and paste the link to the version on your own website.
- Steps 2 and 3: tick the audiences and topics. Both come out of your scan and become tags in the published file.
- Step 4: read the review cards, how many pieces, how many audiences, how many topics, then press "Send Content Box".
- Step 5, the handover pack: download llms.txt and robots.txt, email them to your IT manager, copy the one line of script.
- The live file opens at rankfor.yourdomain/llms.txt, one link per published piece, each one plain text a model can fetch.
- Statistics on a three-month range: clicks that came out of AI conversations, split by which assistant, and the ten pages they landed on.
- The bot donut: every AI crawler that requested one of your pages, ranked by how often it asked.
- Takedown: flip one piece's switch off and it stops being served. Flip it back and it is live again.
- Closing on the project overview: published boxes, content pieces live, last publication.
Shots 7 and 8 are the return leg, and they answer two different questions. The clicks tell you whether a conversation sent a human to your page. The crawler donut tells you which machines came to read it in the first place. A month with crawler requests and no clicks means the models are reading you and not recommending you yet.
What you need
- Marketer, EUR 196 a month. Publishing sits behind that plan.
- A project that already has generated content, personas and topic clusters. The box needs all three.
- A verified domain. Your Rankfor.AI contact sets up the DNS record with you days ahead, because no screen in the app runs that check. Until it resolves, the subdomain answers 403.
- The tracking script live on your website. Statistics collects from the day it goes in, so give it a week or two before you expect a chart worth reading.
Questions people ask
Does this publish to my own website? No. Rankfor.AI hosts the box for you, on a subdomain of your domain. Your own website gets two files and one line of script, once.
Can I see how many people read each piece? No. Per-piece and per-box view counts are missing from the product, and those counters read zero. The traffic numbers are project level: clicks from AI conversations, and crawler requests.
Which assistants show up in the numbers? The chart reads the referrer, so ChatGPT, Gemini and Claude appear as places a conversation happened, and Perplexity can turn up as a traffic source. The scans behind the rest of Rankfor.AI run on Gemini, ChatGPT and Grok.
Is the crawler count complete? No. It comes from the script inside your page, so a crawler that never runs JavaScript never lands in the count. There are no alerts either; you open Statistics when you want to look.
Method: filmed on one project with a verified domain and a live tracking script, Statistics on the 3M range, August 2026.
Open Publishing on your project and run the wizard once. Start with Rankfor.AI.

