What log files tell about your visibility in AI Search

In this presentation, Jérôme shows how log file analysis can uncover your visibility in AI-driven search platforms like ChatGPT, Perplexity, and Claude. The presentation examines how LLM's crawlers interact with websites, how to detect when your content is used in AI-generated answers, and how to track resulting traffic and engagement.

Transcript

2. It’s AI SEARCH Game time !
3. Just because you don’t play, doesn’t mean the game stops. K aaris
4. What is OUR biggest CHALLENGE ?
5. Zero-Clic k Search ?
6. AI SEARCH DATA is our challenge!
7. SEO is a DATA-driven game. But AI Search is a blackbox …
8. There is no AI SEARCH CONSOLE yet…
9. Just clicks… Prompt ? Intent ? Prompt volume ? Visibility share ? Impressions ? CTR ?
10. What happens before clicks ?
11. AI Bots traffic!
12. A LOT OF AI Bots traffic!
13. When AI gives you bot traffic…
14. Make it a LOG FILE ANALYSIS
15. From server log files
16. To log analysis dashboards
17. Same Method Differen t Googlebot log analysis AI bots log analysis
19. Multiple AI platforms
20. Multiple AI platforms Using multiple AI bots
21. Multiple AI platforms Using multiple AI bots For multiple crawl purposes
22. How do AI platforms need these BOTS ?
23. Model training knowledg e Web search tool (Grounding) PROMPT
24. TRAINING 1 BOT SEARCH 2 BOT USER 3 BOT 2 Teams / 3 Players
25. TRAINING 1 BOT SEARCH 2 BOT USER 3 BOT Knowledge team
26. Training Bots Crawl your website to SCRAPE CONTENT that MAY be used in LLM training.
27. User agents (bot names) GPTBot, ClaudeBot, CCBot, Bytespider
28. No crawl patterns
29. Crawling every page they can find AI Training Bots GoogleBot
30. Is it useful to track training bots ?
31. Most collected data STAY on the bench .
32. Collected data is filtered and preprocessed. Data-cleaning pipeline Step What happens Crawling Bots collect raw web content Filtering Remove spam, duplicates, and low-quality data Classification Identify topics and content types Sampling Keep diverse, balanced examples Pre-processing Format and prepare data for training Curation Select the final subset used in the model
33. NO real INSIGHTS Training bot crawl ≠ Model training
34. Search OFF Search ON ChatGPT queries study (Semrush Oct/Nov 2024)
35. Prompt intent study (OpenAI Sept 2025)
36. GPT-5 favors Search over Knowledge (Dan Petrovic Aug 2025)
37. TRAINING 1 BOT SEARCH 2 BOT USER 3 BOT Grounding team
38. AI Search bots AI User bots Grounding Dream Team
39. PROMPT with Search WEB SEARCH CONTENT SCRAPING RESPONSE with citations & sources GROUNDING PROCESS 1 2 3 4
40. 2 WEB SEARCH The system queries a search engine and/or a custom index.
41. AI Search bots Crawl your website for INDEXING or IMPROVING SEARCH results.
42. User agents (bot names) OAI-Searchbot, PerplexityBot, Claude-Searchbot
43. AI Search bots are similar to Googlebots BUT…
44. They don’t crawl as much as Googlebot
45. Crawl frequency per page is 1 visit per day OAI-SearchBot PerplexityBot
46. The crawl pattern favours new pages OAI-SearchBot PerplexityBot
47. AI Search Bots only show one side of the court…
48. PROMPT with Search WEB SEARCH CONTENT SCRAPING RESPONSE with citations/sources GROUNDING PROCESS 1 2 3 4
49. 3 CONTENT SCRAPING The system sends a User Bot to crawl pages selected from the search results.
50. AI User bots Crawl your website to FIND THE ANSWER to a USER PROMPT in REAL-TIME.
51. User agents (bot names) ChatGPT-User, Perplexity-User, Claude-User, MistralAI-User
52. AI User bots let you enter the court…
53. AI User bots crawl show how often AI Chatbots try to use your content
54. ChatGPT-User weekly crawl behaviour
55. ChatGPT-User daily crawl behaviour
56. User bot activity ≃ AI Chatbot prompt activity
57. User bot hits ≃ Impressions
58. But it DOES NOT always mean CITATION…
59. Poor content NEVER plays
60. User bot crawl blocked
61. Client-Side JS User bot crawl blocked
62. Client-Side JS Paywall User bot crawl blocked
63. Client-Side JS Paywall Status code User bot crawl blocked
64. ChatGPT uses a caching system
65. ChatGPT follows redirect
66. How to use this data as a key VISIBILITY METRIC ?
67. Filter by STATUS CODE 200 and 304 for more accurate data
68. 2 metrics to report AI Search visibility How many visits from AI bots How many pages crawled by AI Bots
69. Measure AI Search visibility progression
70. Identify top crawled pages by AI Bots
71. Calculate custom ratio (Clicks / Bot hits) * 100 (Crawled / known pages) * 100
72. Actionable STATS that WIN the GAME!
73. Citation missed opportunities
74. Crawled pages that never get clicks
75. Pages never crawled by AI bots but with SEO visits in Google
76. Segmentation analysis By language
77. Segmentation analysis By article topic
78. Segmentation analysis By page type
79. Crawl X logs analysis ChatGPT-User visits by content size
80. Log analysis X Search Console ChatGPT-User visits by GSC avg position
81. And more to EXPLORE
82. Other highlights in your logs…
83. AI ChatBot Updates
84. Model updates impact on crawl behavior
85. ChatGPT Search update June 13, 2025
86. GPT 5 release August 7, 2025
87. Robots.txt ?
88. ChatGPT-User stopped crawling robots.txt
89. LLMs.txt ?
90. AI Bots don’t crawl this file yet…
91. Using log files to track AI Search traffic
93. No sampling No tracking loss Status code 1 2 3
94. LOG ANALYSIS From bench to MVP
95. Ready to PLAY ?
96. Scan to Connect /jeromesalomon OVERTIME! Book a demo and let’s keep the conversation going. The game continues…