AI crawlers in your server logs: GPTBot, ClaudeBot & co.
How to find GPTBot, OAI-SearchBot, ChatGPT-User, ClaudeBot and PerplexityBot in your logs, verify their IPs, and read training versus live-fetch patterns.
What ChatGPT, Gemini and Perplexity recommended - and cited - across law firms, clinics, ecommerce, hotels and six more industries in the US, UK, Canada, Spain and Turkey. 79% of cited domains appeared exactly once, ChatGPT and Gemini named the same #1 brand in 8.5% of questions, and Wikipedia made up 9.5% of ChatGPT's citations but none of Perplexity's.
Every month we freeze a snapshot of how AI assistants answer the questions buyers actually ask - "best divorce lawyer in Toronto", "dental implants clinic in Madrid", "best project management software" - and publish it as the Zene AI Visibility Index. The October 2026 edition covers 617 engine answers to 210 questions across 10 industries and 5 countries (US, UK, Canada, Spain, Turkey), from ChatGPT, Gemini and Perplexity. The answers were captured in July 2026; every number below comes from that frozen edition, and the per-industry answers and CSVs are on the research pages.
ChatGPT and Perplexity cited 1,884 sources across 1,106 different domains. The ten most-cited domains together account for just 14% of citations, and 869 domains - 79% - were cited exactly once.
That is the opposite of classic search, where a handful of domains own page one. For a brand it means there is no single site to "get onto": AI answers are assembled from whatever specific pages answer the specific question - a local directory, one comparison article, a niche review site. The work is finding which pages your questions pull from, not chasing one magic placement.
An answer named 6.4 brands on average. Comparing ChatGPT and Gemini on the same question, the brands they named overlapped by only 16%, and they picked the same #1 brand just 8.5% of the time (199 question pairs).
Part of that is mechanics: in this edition ChatGPT searched the web and cited pages, while Gemini answered from the model alone. But the practical lesson holds either way - being the top pick in one engine tells you almost nothing about the others. We wrote about the reasons in why AI engines recommend different brands; the index puts a number on it.
Wikipedia was the single most-cited domain (92 citations, in 8 of 10 industries) - and every one of them came from ChatGPT, where it made up 9.5% of all citations. Perplexity didn't cite Wikipedia once in this edition.
The reverse is true for community and video: Reddit and YouTube were 5.4% of Perplexity's citations and absent from ChatGPT's. If your category's questions run through Perplexity, the threads and videos about you are part of your visibility; for ChatGPT, a factual, well-sourced Wikipedia presence (where you genuinely qualify for one) carries more weight.
Below the cross-industry regulars, the most-cited sources were niche:
For a local or vertical business, the directory or comparison site your industry's answers lean on is usually the highest-leverage place to be listed accurately - and it is different in every vertical.
Hotel questions drew 1.1 citations per answer, against 4.5 for law firms. Assistants recommended plenty of hotels (441 brand mentions) but rarely showed their work - the names come mostly from the model's general knowledge and big travel brands rather than from cited pages. For hotels that makes the classic signals - consistent listings, reviews, a clear entity across booking sites - more important than any single article.
For each industry and country we ask a fixed set of buyer-intent questions in the local language, record every engine's full answer, extract the brands named (in order, with sentiment) and the sources cited, and freeze the result as a monthly edition. Counts here are distinct citations per answer, aggregated across the 42 industry-country datasets in the edition. Small datasets (SaaS and cybersecurity are US-only this month) are directional, not definitive.
Next edition: November 2026, the first one with a month-over-month comparison - which brands gained and lost share of AI answers in each industry.
Want the same view for your own brand - which engines name you, who they name instead, and which pages they cite? Run a free audit, or browse the full October 2026 index.
A monthly, frozen snapshot of how AI assistants answer high-intent questions in an industry and a country - which brands they name, in what order, and which sites they cite. Every edition is published at tryzene.com/research with the underlying answers and a CSV per industry and country, so the numbers can be checked rather than taken on trust.
They retrieve and weigh different evidence. In this edition ChatGPT's answers were web-grounded and cited sources, while Gemini answered from the model alone; ChatGPT and Gemini shared only 16% of the brands they named for the same question and agreed on the #1 pick 8.5% of the time. Being visible in one engine says little about the others, which is why tracking has to be per engine.
No single site dominates. Across 1,884 citations, the ten most-cited domains accounted for 14% and 869 of 1,106 domains were cited exactly once. The cross-industry regulars were Wikipedia (mostly in ChatGPT answers), YouTube and Reddit (mostly in Perplexity answers); within an industry, niche directories and comparison sites mattered more - WhatClinic for clinics, LawyersBay for law firms, Clutch for agencies.
Put this guide into practice - get your free visibility score in minutes.
