For AI agents: a documentation index is available at /llms.txt
AXO · Observed
See the AI agents that visit your site
WorkPods runs the Observed layer of AXO (agent experience optimization): we show your marketing team which AI agents visit your site and which pages they fetch. Google Analytics 4 (GA4) depends on browser JavaScript that AI bots do not run, so your server logs are where those visits show up.
GA40agent visits: agents run no JavaScript
Your logs14,820agent requests this month
| Agent | Purpose | Page | Status |
|---|---|---|---|
| GPTBot verified | Training | /pricing | 200 |
| ChatGPT-User verified | User fetch | /book-demo | 200 |
| ClaudeBot verified | Training | /llms.txt | 404 |
| PerplexityBot verified | Search index | /compare | 403 |
| Google-Agent verified | Agent browser | / | 200 |
Checked against each operator's published IP ranges. No IPs stored.
01
See which AI agents visit your site
GA4 misses AI crawlers and assistant fetchers because they read HTML or Markdown straight from your server and never run browser scripts. When a buyer asks an assistant to check your pricing or security page, that fetch leaves no trace in your analytics dashboard.
To capture those visits without adding page scripts, WorkPods reads Cloudflare (Worker or Logpush), Vercel log drains, or uploaded nginx and Apache logs. We check each bot request against its operator's published IP ranges, then group visits by operator and by purpose: training, search index, user fetch and agent browser.
02
Turn agent visit logs into marketing actions
Verified visit data gives your web team a fix list and gives leadership a plain report on AI reach. A CMO or growth lead does not need to read server lines to use it.
It supports four decisions:
- See which pages AI tools read before buyers arrive: user-fetch requests show which pricing and product pages assistants pull, supporting your GEO work.
- Fix pages that serve errors or blocks to agents: when AI fetchers hit server errors or blocks, those URLs move into fixes for agent-ready websites.
- Spot priority pages agents never fetch: when core guides get no fetches in covered logs, add links in llms.txt (a short file listing key pages).
- Report AI reach alongside GA4: setting human GA4 sessions next to verified search-index and user-fetch visits can help explain why demand holds up when search clicks fall.
03
Review a weekly agent summary alongside GA4
A weekly summary gives marketing leaders a short readout of agent activity, which pages were fetched and what blocked agents. The table below is a hypothetical example of how we format a weekly review.
It is not a real client result.
| Metric category (hypothetical example) | Example weekly observation | Marketing or web action |
|---|---|---|
| Agents by operator and purpose | Search-index and user-fetch visits Verified via IP check; Mismatch spoofed bots filtered out | Focus reporting on Verified search-index and user-fetch visits that shape buyer answers |
| Top pages fetched by user-fetch bots | Pricing overview, enterprise security page and comparison guide fetched most often | Keep copy, structured data (labels telling machines what a page is) and pricing current on these pages |
| Errors and redirects served to agents | Product comparison URLs returning blocked responses or redirect chains to Verified search bots | Update bot rules and clean up redirects so assistants stop dropping those pages |
| Unfetched priority pages and log coverage | Core integration pages show no fetch observed in covered logs across 7 days received | Add missing integration links to llms.txt and main site navigation |
04
How we measure it
We check every agent request in the logs you send and record how your server responded, before and after each fix. Each visit gets a verification state, and every report shows how many days of logs it covers.
- For Observed agent visits, each log entry is Verified inside published IP ranges, Mismatch outside them, or Unverified without published ranges.
- Coverage in days of logs received is always shown, and "no fetch observed in covered logs" is never stated as "never fetched".
- Each AXO check on your page responses and site files is scored Pass, Partial, Fail or Not applicable. A missing optional file such as llms.txt is Not applicable, never a fail, and only a present, broken file gets Fail.
WorkPods never stores visitor IP addresses. We discard each IP right after the operator range check and keep only daily totals, plus cleaned log batches for 14 days. Client data stays in WorkPods' own infrastructure with role-based access and an approval log, and a data processing agreement (DPA) is available on request; details are on trust and data handling. You can see what an operator finds on your public site first with a free GEO and AXO audit, delivered within 3 working days.
Questions
Questions buyers ask
Why can't Google Analytics 4 see AI agent visits?
GA4 counts a visit only when its JavaScript tag runs in a visitor's browser. An AI crawler or assistant that fetches your pricing page or llms.txt file asks for the HTML or Markdown and runs no JavaScript. The tag never fires, so GA4 records nothing for that agent visit, and only your server or CDN logs hold it.
How do you tell real AI agents apart from fake scrapers?
We check each request's IP address against the ranges AI operators publish, because the user-agent name alone proves nothing. A scraper can copy a well-known bot's name into its request header. We mark each visit Verified inside the operator's ranges, Mismatch outside them, or Unverified when the operator publishes no ranges.
What is the difference between training crawlers and user-fetch bots?
Training crawlers collect text in bulk for future language models. Search-index bots crawl pages so AI search engines can ground their answers. User-fetch bots and agent browsers request a page in real time because a person just asked an assistant about your brand. For marketing reports we focus on search-index and user-fetch visits.
Which server and CDN log sources do you support?
We read Cloudflare (through a Worker or Logpush), Vercel log drains, and uploaded nginx or Apache server logs. Each report states how many days of logs it covers. The logs come from your server or edge network, so connecting them adds no JavaScript to your pages and does not slow them down.
Do you store visitor IP addresses from our logs?
No. WorkPods never stores visitor IP addresses from your logs. Each request is checked against published IP ranges and marked Verified, Mismatch or Unverified, and the IP address is then discarded. What we keep is daily totals, plus cleaned log batches for 14 days, after which only the totals remain.
Do we need to share server logs for the free audit?
No. The free audit needs no server logs, analytics access or CMS credentials. The form asks only for your work email and website. An operator checks your public pages, site files and how six AI answer engines describe you, and sends the report within 3 working days. Log analysis starts once you choose to connect logs.