Session data: March 3–10
| Metric | Value | | --- | --- | | ChatGPT-User visits | 681 | | Reconstructed sessions | 168 | | Content sessions (non-index) | 127 | | Pages fetched | 587 | | Avg pages per content session | 4.6 | | Multi-turn sessions | 38 (30%) | | Unique questions identified | 109 |
The session reconstruction is imperfect. Heuristics infer behaviors. The patterns that emerge are coherent enough to be meaningful.
From previous weeks, the number of ChatGPT-User visits has gone down. The counting method was refined.
The signal: ChatGPT-User in CloudFront logs
ChatGPT-User is a documented OpenAI bot. ChatGPT-User fetches web pages during response generation. ChatGPT-User is distinct from GPTBot. GPTBot indexes content for training. ChatGPT-User is distinct from OAI-SearchBot. OAI-SearchBot builds the retrieval index.
ChatGPT-User shows up at response time. ChatGPT-User shows up when a human uses ChatGPT. ChatGPT-User is not associated with the ChatGPT API usage in this context. ChatGPT-User queries the AI Site in real time.
The AI Site is hosted on CloudFront. Every ChatGPT-User request is stored in logs with a timestamp. Every ChatGPT-User request is stored in logs with a URI. Every ChatGPT-User request is stored in logs with an IP hash. Those 3 pieces of information are used to reconstruct multi-turn ChatGPT sessions.
The result is a session model. Each session has one or more turns. Each turn corresponds to a fetch event. During a fetch event, ChatGPT-User pulls a set of pages to respond to something. Page contents are analyzed to infer what the user was asking about at that turn.
Finding 1: ChatGPT-User fetches multiple pages per turn
A naive model of AI retrieval assumes one query maps to one fetch. The naive model assumes the bot finds the best page. The naive model assumes the bot reads the page. The naive model assumes the bot answers.
The log data does not support the naive model. In reconstructed turns, ChatGPT-User on average pulls 4.6 pages. The 4.6 pages are pulled in a tight burst. After the burst, ChatGPT-User moves on.
Pricing questions generated the most fragmented fetch patterns. Across the week, four near-duplicate Q&A pages covering pricing were fetched a combined 47 times.
| Page | Visits | | --- | --- | | what-pricing-plans-are-available-for-genymotion | 18 | | what-are-genymotion-s-pricing-options | 15 | | what-are-the-pricing-options-for-genymotion | 7 | | what-are-the-costs-for-using-genymotion-saas | 7 |
The same pattern appeared for macOS compatibility. The same pattern appeared for Google Play installation.
macOS compatibility involved 3 pages. macOS compatibility involved 38 combined visits. Google Play installation involved 4+ pages. Google Play installation involved 30+ combined visits.
One interpretation is presented. ChatGPT-User is verifying and consolidating across sources. ChatGPT-User is not just reading the first relevant result.
If the interpretation is right, it justifies an aspect of the AI site design. Multiple facets of a topic are covered by distinct Q&A pages. Those distinct Q&A pages exist due to variations in the way users ask questions in the chatbot.
This is not redundant information. Providing multiple facets may provide complementary information. Providing multiple facets may validate information. This is what the bot may be looking for.
Finding 2: 28% of sessions hit the index
40 of 168 sessions fetched only /index.html.
Those sessions stopped after fetching /index.html.
No Q&A pages were fetched in those sessions.
No content pages were fetched in those sessions.
No further navigation page queries were visible.
The previous index was an infrastructure page. The previous index listed API endpoints. The previous index listed content counts. The previous index listed navigation links.
ChatGPT-User does not use JSON APIs.
ChatGPT-User GETs HTML pages.
When ChatGPT-User arrived at the index, /index.html, it had insufficient signal.
The session ended on the AI Site side.
No further turns were visible.
No further page queries were visible.
This ended as if the bot had moved on.
An adaptation was made to the index page. The adapted index page opens with a product description. The adapted index page includes a topic directory. The topic directory includes inline descriptive summaries. The inline descriptive summaries give the bot enough context to proceed.
Improvement is scheduled for checking in the next weeks. The check is for AI Site bot performance.
Finding 3: 30% of sessions are multi-turn
Nearly 1/3 of sessions involved a second or third fetch cluster. The second or third fetch cluster can be attributed to the same session.
A meaningful conversation can be followed. The conversation happened in ChatGPT. The conversation did not happen in the site’s own chatbot.
| Date | Turns | Pages | Duration | Fetch pattern | | --- | --- | --- | --- | --- | | Mar 3 | 4 | 11 | 248s | Web emulator → installation → pricing → user guide | | Mar 4 | 3 | 21 | 95s | macOS compatibility → pricing → Play Store setup | | Mar 6 | 4 | 6 | 637s | SaaS templates → template count → index → user guide | | Mar 7 | 2 | 9 | 1,377s | Linux install + KVM errors → ARM transition |
The March 4 session fetched 21 pages. The March 4 session occurred across 3 turns. The March 4 session duration was 95 seconds.
The March 7 session had a 23-minute gap between turns. The gap indicates a shift away by the user. The gap indicates a return with a follow-up question.
Multi-turn sessions reveal sequential nature. Citation tracking cannot reveal the sequential nature. An example citation tracking metric is Share of Citations.
Fetch patterns show natural progressions. Progressions can move from compatibility to pricing to setup. Progressions can move from troubleshooting to related technical questions.
A question is posed about the probability of recommendations. The question is about whether the AI will recommend the site.
Get This for Your Site
ROZZ builds this infrastructure automatically. ROZZ provides AI site. ROZZ provides Q&A pages from a chatbot. ROZZ uses Schema.org markup on every page. ROZZ derives session analytics from weekly log analysis.
$997 /month | 168 sessions reconstructed in one week
→ Book a call | → See how it works | → rozz@rozz.site
Data source and reconstruction notes
Data source is CloudFront access logs for rozz.genymotion.com. The data period is March 3–10, 2026.
Session reconstruction is based on IP hash grouping and timing heuristics. Bot classification is based on User-Agent strings.
Author
Author: Adrien Schmidt, CEO, ROZZ
Serial tech entrepreneur with 10+ years experience building AI systems. AI systems include Aristotle (conversational AI analytics). Products include Aristotle for conversational big data analytics chatbot. Products include products for eBay and Cartier. Previous founders include Squid Solutions. Built AI products like Aristotle. Built an AR jewelry try-on device for Cartier.
Publication date and data period
March 10, 2026 | Data period: March 3–10, 2026
rozz@rozz.site | © 2026 ROZZ. . All rights reserved.