A number of brand-new records and analyses of agentic website traffic fads highlight why publishers are constructing updated approaches around AI visibility, in an internet environment currently comprised of more bots than humans.

Agentic website traffic is surging

A report from DataDome published recently programs AI agent traffic grew 45 % in Q 2 2026 DataDome’s network of over 400 firms tape-recorded 17 7 billion AI agent requests in April-June 2026, a 45 % rise from Q 1’s 12 2 billion. June had 6 6 billion AI representative requests alone.

Meta currently generates most of AI agent traffic on DataDome’s network, both from its training and cloth crawlers Meta’s training spider grew 74 % from Q 1 to Q 2, while its RAG crawler expanded faster, up 163 % in the very same period.

“We’re entering a new phase of the AI internet,” claimed DataDome vp Jérôme Segura. “Meta is moving from ‘scrape when for training’ to continual indexing for real-time AI answers. As AI web traffic ends up being extra persistent, authors need to establish guidelines for crawling and indexing like they have been doing for Google. It’s important to comprehend which representatives accessibility content, how typically, and what they get in return.”

At the same time, a July record from Decodo examined Cloudflare information and found AI-driven website traffic grew by roughly 187 % in 2025, around 8 times faster than human web traffic over the same period. Automated systems now produce 57 4 % of web demands, compared with 42 6 % from people, according to Decodo’s analysis. AI agent traffic expanded by around 7, 851 % year over year.

Much more internet sites are developing for representatives

To get ready for the agentic internet, more companies are developing agent-readable versions of their internet material to make it less complicated for AI agents to gain access to and understand. Markdown and LLMs.txt– a specific kind of markdown file– are web styles that remove HTML information to offer representatives content and metadata, without the surrounding web page design.

AI discovery business Originality.ai released a record and dashboard last month that tracks the growth of AI web standards like LLMs.txt.

The report located that LLMs.txt circumstances expanded from 4, 088 in June 2025 to 36, 120 by May 2026, an 8 8 X rise across the 3 million sites it evaluated. Originality.ai determined 38, 980 websites adopting web standards like llms.txt, llms-full. txt, and ai.txt.

Nonetheless, some publishers may lead the video game right here. DataDome’s report located that LLMs.txt has gotten to just 3 2 % adoption among authors in July, listed below the 7 9 % on the open web.

And Ahrefs’ server-log research study discovered 97 % of LLMs.txt data received no requests in May 2026, according to Originality.ai’s report. Jon Gillham, Chief Executive Officer at Originality.ai, stated this means markdown pages may not be the trick to enhancing AI visibility. (Google’s very own GEO assistance is that LLMs.txt is not necessary for generative AI search optimization )

“The 8 8 [times] growth in LLMs.txt fostering shows that publishers are searching for some control over how AI interacts with their content. But fostering isn’t usage. If 97 % of these documents are never ever requested, this isn’t an AI presence method yet. It’s an affordable bet on an agent-driven future which could include llms.txt,” Gillham stated.

AI visibility is a challenging video game to play

An additional indication that AI presence is a hard video game to play– regardless of it becoming a metric of growing value for some publishers– is the fact that it’s not easy to get pointed out or pointed out in an AI response.

Webflow, an aesthetic web advancement system, produced an” AEO Maturity Index that evaluates a company’s search engine optimization and GEO efficiency. It assessed 2, 000 websites, considering their published content, technological GEO readiness, brand name authority and dimension capabilities. It found that the median firm showed up in 16 % of relevant AI answers, and showed up in a citation with a web link just 6 % of the moment, according to Person Yalif, chief evangelist at Webflow. Problems with the fundamental fundamentals of search engine optimization were normally culprits for underperforming GEO, he said, such as busted web links, missing metadata, and stale material.

“We are in early days in a brand-new tool,” Yalif stated.” [Publishers] can, at [their] core, answer inquiries, make it really simple for the LLMs to eat their web content, show up in a number of places, and gauge the right things.”

Webflow developed a rubric called the Webflow AEO Maturation Version in these four classifications. The even more prepared a business was for GEO, the higher the mention and citation prices they were seeing.

Huge companies had a tendency to execute better, according to Webflow’s information, with two-times greater reference rates, 60 % higher citation prices and 80 % greater share of voice in AI search.

AI recommendation traffic still weak

New information also illustrates why authors appear to be shifting far from seeing AI recommendation traffic as a growth opportunity.

DataDome’s record located that ChatGPT-User, which fetches live pages to respond to real-time questions, went to internet sites 6 % less in Q 2 than it performed in Q 1

Yet, it’s not all bad news for publishers. People making use of ChatGPT clicked through to websites 17 % even more, with ChatGPT accounting for upwards of 80 % of all AI-driven recommendation traffic.

DataDome’s Segura claimed this indicate why publishers are relocating past covering AI crawler policies. Reuters and Time, as instances, have actually lately begun blocking all AI bots by default however developed whitelists of accepted robots to accessibility web content on their sites.

“Not all AI representatives are the same,” claimed DataDome’s Segura. “The inquiry requires to advance past merely allowing or obstructing AI representatives to establishing which ones must be relied on and why. The representatives worth focusing on aren’t just the ones striking your web servers hardest, but additionally the ones really sending visitors. The publishers finest positioned for the agentic web are those that build policies around a representative’s purpose, identification, and company worth, instead of treating every AI agent the very same.”

Some authors are still attempting to obstruct robots

Mentioning crawler blocking, more new publishers are moving to obstruct AI bots– even if the methods to do so aren’t iron-clad. A report from internet information firm HasData found that thus far this month, over fifty percent of news authors were obstructing a minimum of one AI spider in robots.txt.

Of the 10, 894 domain names that were assessed, 56 4 % of information publishers now block a minimum of one AI crawler, contrasted to 10 3 % of the open internet. GPTBot especially is banned by 50 5 % of authors, contrasted to 7 9 % of the web.

However bot barring doesn’t always work. Of the sites that declared a GPTBot ban and were live-tested, 39 5 % offered GPTBot a web page anyhow, HasData found.

“Long-term, publishers remain in a ‘change-or-die’ area due to the fact that their security only works with courteous robots. So, most likely they are losing their AI search visibility, and at the very same time, not quiting the web traffic they fear most,” stated Roman Milyushkevich, CTO and Chief Executive Officer of HasData. “It can be exceptionally difficult to distinguish agent traffic from internet browser web traffic. So authors ought to think of correct access bargains and money making that would benefit both of these classifications as opposed to basic wall surfaces.”