Developing
A vendor press release with limited methodology. The figures are BrightEdge's.
In 60 seconds
- BrightEdge reported on April 8, 2026 that AI agent requests had reached 88% of human organic search activity.
- Agents were about 15% of total website traffic, 95% of that from OpenAI.
- 77% of companies focus on blocking training agents; 21% address search agents and 38% user-facing agents. Only 19% of sites have directives for ChatGPT-related bots.
- BrightEdge projects agent activity will pass human search by the end of 2026; this is a projection.
Impact Layer
- What changed
- A search-software vendor put a number on agent traffic and on the policy gap.
- Why it matters
- If agents act for customers, blocking them can block sales; if they are not managed, content may be used in ways owners did not intend.
- Who is affected
- CMOs, SEO and web teams, and whoever owns robots.txt and bot management.
- Martech stack area
- AI, SEO & Search
- Stack impact
- Separate your policy for training, search and user-triggered agents, and monitor which bots reach your site.
- Business impact
- Moderate Methodology details were limited in the announcement; the projection is BrightEdge's.
- Related technologies
- What to watch
- Independent measurements of agent traffic, and how Cloudflare-style defaults change the numbers.
BrightEdge, which sells search and content software, published data on April 8, 2026 on how AI agents use the web.
The figures
- AI agent requests have reached 88% of human organic search activity.
- Agents are about 15% of total website traffic, and OpenAI drives 95% of that share.
- BrightEdge projects agent activity will surpass human-driven search by the end of 2026.
- Only 19% of websites have directives specifically for ChatGPT-related bots. 77% of companies focus on blocking training agents, while 21% address search agents and 38% manage user-facing agents.
CEO Jim Yu is quoted: “If you block or fail to optimize for these agents, you’re not blocking bots, you’re blocking customers.” BrightEdge says methodological details were limited in the announcement.
Our read
The number to remember is the policy gap, not the headline ratio. Most organizations decided about training bots and left the rest unmanaged. A separate rule for each type is easy to write: block training if you object to it, allow search crawlers so you can be found, and decide on user-triggered fetching on its merits. The 88% ratio is the vendor’s measure and the projection is a forecast; treat both with care and compare with your own server logs.
Sources
- GlobeNewswire: BrightEdge data, AI search is reaching a tipping point (opens in a new tab)
Source of every figure and the quotation.