Akamai analysis of AI robot activity by examining application layer traffic from its bot management tools.
Commerce Has Attracted the Most AI Bot Traffic Currently 48%. Media, which includes publishing, video, social media and broadcasting, comes in second. 13%.
Publishing companies represent 40% of all AI bot activities in the media, before broadcast and OTT to 29%.
OpenAI generated the most AI bot traffic from media companies, with 40% of its media requests going to publishing companies. This is partly because OpenAI runs multiple robots. GPTBot manages the training, OAI-SearchBot powers the AI search and ChatGPT-User retrieves content in real time.
Meta and ByteDance were the second and third largest operators. Anthropic and Perplexity round out the top five at lower volumes.
Why Akamai says Fetcher bots are the biggest concern
The report groups AI robots into four types based on their behavior.
Training crawlers and scavengers accounts for most of the AI bot activity Akamai has seen in the media, including publishing. Training bots collect content to create language models. They invented 63% of AI bot activity targeting media in the second half of 2025.
Retrieval bots retrieve specific pages in real time when someone asks an AI chatbot a question. They invented 24%and the edition represented 43% of this recovery activity.
Akamai says scrapers are the most immediate revenue concern, even though crawlers generate more total traffic. When a retrieval bot retrieves an article to respond to a chatbot query, the user gets the information without visiting the publisher’s site.
How publishers respond
It should be noted that Akamai sells bot management tools and the report’s recommendations are aimed at its own products and partners.
The most common responses among Akamai customers are deny (outright blocking requests), tarpit (keeping connections open to waste bot resources), and delay (adding a pause before responding). An anonymous publisher chose tarpiting over blocking, vetted 97% of AI bot requests, and kept the door open for possible licensing deals.
The report argues against blanket blocking, saying some AI companies are willing to pay for access to content and blocking all bots removes that option.
Looking to the future
The main takeaway from the report is the distinction between crawlers and retrieval bots. Blocking a training crawler can influence how your content helps build future AI models. Blocking a crawler affects whether your content currently appears in AI responses.
Featured Image: pico de gallo/Shutterstock
