Eleven Machine-Readable Resources in 92 Seconds
A GPTBot-classified client requested AS400Software.com's catalog and ten JSON datasets in 92 seconds. The homepage was the recorded referer, and those eleven resource URLs appeared on the homepage only inside JSON-LD.
Audio Summary
Listen to the Summary
This 71-second recording is a summary, not the full proof entry.
Read by Kodi, AI co-architect
Read transcript
Here's what happened.
On the AS400Software.com website, we put information in a part of the webpage that people normally don't see. It's called JSON-LD. Think of it like leaving a map for computers.
That map pointed to 11 different data files on the website. Those files were not linked anywhere people could click on the homepage. They were only listed inside that hidden computer-readable information.
Then we saw something interesting.
A visitor that our system identified as GPTBot came to the homepage. Right after that, it requested the data catalog and 10 of those JSON files. It did all of that in just 92 seconds, and our server recorded the homepage as the page it came from.
So we know the bot visited the homepage, and then it visited 11 resources that were listed in the homepage's structured data.
What we cannot prove is exactly how the bot found them. We can't see inside the bot and say, "Yep, it read our JSON-LD and followed the links."
But we can see what happened on our server.
We put out a map for bots. A bot showed up. And then it visited almost everything on the map.
That's the receipt.
On August 10, 2026, AS400Software.com recorded a compact request sequence that is hard to ignore. A client classified as GPTBot requested the machine-readable catalog and ten JSON datasets in 92 seconds. The homepage was recorded as the referer.
A fresh source audit found all eleven resource URLs inside the homepage JSON-LD. None appeared in the visible homepage links. Every resource still returns HTTP 200 and valid JSON. Together, the declared datasets cover 369 records.
The larger retained log sample contains 49 successful endpoint requests across 12 JSON and AI resources from four crawler fingerprints. The 92-second sequence is the cleanest receipt inside that activity because the request order, referer, declarations, and live resources can all be checked independently.
Evidence
- Date: August 10, 2026 - Client classification: GPTBot-classified - Sequence: catalog plus 10 JSON datasets - Elapsed time: 92 seconds - Recorded referer: AS400Software.com homepage - Visible homepage links to those resources: 0 - Homepage JSON-LD declarations for those resources: 11 - Current HTTP and JSON validation: 11 of 11 passed - Records represented by the declared dataset layer: 369 - Broader retained endpoint activity: 49 successful requests across 12 resources from four crawler fingerprints
Methodology
The receipt combines live warehouse request records with a fresh source inspection and live endpoint validation. Requests were limited to GET responses with successful status, final bot or AI-crawler classification, active fingerprints, approved crawler purposes, and Krisada's default IP exclusion. The homepage source was checked to identify where each requested URL appeared. Each declared resource was fetched again and decoded as JSON.
This is a reconstructed observation, not a controlled experiment. A referer can identify the previous page without revealing whether the client followed a JSON-LD declaration, read another bot file, or arrived with prior knowledge of the URL.
Outcome
Supported observation: a structured-data declaration sat immediately upstream of a real bot-resource request sequence.
Not demonstrated: JSON-LD parsing, semantic understanding, indexing, citation, training use, ranking influence, or any other downstream effect.
The receipt now has a controlled follow-up instead of being stretched into a conclusion it cannot carry.
See More Evidence
More SEO proof from live websites, each one with the numbers attached.