Picture the moment you ask ChatGPT or Perplexity something that needs up-to-date information. Behind that single question, the assistant fires off requests to the pages most relevant to it, reads their text, and builds an answer from what it finds. Each of those requests is a Live Search Agent: a bot dispatched in real time, on behalf of one user's question, to read one specific page. It is not a training crawler gathering content in bulk for later, it does not run any JavaScript on the page, and because of that it never shows up in standard analytics. That combination, real-time, intent-driven and invisible, is what makes it matter to publishers.
How a Live Search Agent differs from a training crawler
The two bot types look similar in a server log but do completely different jobs. A training crawler, such as the bots used to gather data for model training, visits broadly and infrequently to build a static snapshot of the web. A Live Search Agent visits narrowly and on demand, fetching only the pages relevant to a live query, often within seconds of a user asking. Training crawlers feed the model's long-term memory. Live Search Agents feed a single, specific answer right now. Because the Live Search retrieval is tied to an active user question, it is far closer to the moment of intent, and it is the category growing fastest as AI assistants become a default way people find information.
Why Live Search Agents matter to publishers
Live Search Agents read publisher content and generate no revenue from it. A human visit triggers analytics, an ad auction, affiliate links, and any paywall logic, all of which run in JavaScript. A Live Search Agent makes an HTTP request, receives the HTML, parses the text, takes what it needs, and leaves. None of the JavaScript fires. The publisher's analytics do not record the visit, the ad auction never runs, and there is no record the retrieval happened. In live publisher accounts, Live Search is frequently the single largest category of AI traffic by volume, which means a large and growing share of a publisher's real audience is both invisible and unmonetised.
Why standard analytics cannot see them
Google Analytics and most client-side measurement tools rely on JavaScript executing in the visitor's browser. Live Search Agents do not run JavaScript, so they never appear in those dashboards. Publishers making editorial and commercial decisions on client-side analytics are therefore working from incomplete data, missing the fastest-growing segment of their traffic entirely. Seeing Live Search Agents requires measurement at the server or CDN layer, where the raw request is visible before any page render.
What a Live Search Agent retrieval reveals
A retrieval is not just a bot visit. It is a signal. The agent fetched a particular page because a user asked a particular question, which means each retrieval carries information about what the audience is researching and how close they are to a decision. Clustered across a domain, these retrievals reveal named intent segments, the buyer-journey stage behind them, and the commercial value of the audience reaching that content through AI. Read this way, Live Search Agent traffic is among the most valuable audience data a publisher has, because it is tied directly to active purchase research.
How Live Search Agents are monetised
Because Live Search Agents read text and skip JavaScript, they cannot be monetised by display ads or any client-side mechanism. The monetisation has to live in the content layer, in the text the agent reads, inserted server-side before the agent processes the page. blankspace operates here. It detects Live Search Agent retrievals at the CDN edge, runs a sub-50ms auction, and injects a relevant, accurate brand fact into the content as editorial context, invisible to human readers, with revenue attributed the same day. This turns a previously invisible, unmonetised visit into measurable income.
How Live Search Agents are identified and verified
Identifying a Live Search Agent reliably matters, because a user-agent string can be spoofed. The user-agent header is the weakest signal. More reliable methods include checking the request against the published IP ranges of the AI company, confirming reverse DNS, and verifying cryptographic signatures where the agent supports an authentication standard. Accurate identification is what lets a publisher attribute traffic to the correct owner, apply the right policy, and monetise genuine retrievals without being misled by impersonators.
Frequently asked questions
Is a Live Search Agent the same as GPTBot or a training crawler?
No. Training crawlers like GPTBot gather content in bulk to train models. A Live Search Agent fetches a specific page in real time to answer a live user question. They serve different purposes and have different commercial implications.
Why does a Live Search Agent not show up in Google Analytics?
Because it does not execute JavaScript. Client-side analytics depend on JavaScript running in the browser, so any agent that only requests and parses the raw HTML is invisible to them. Server-side or CDN-level measurement is needed to see Live Search Agents.
Which AI assistants use Live Search Agents?
The major assistants all retrieve live content for queries that need current information, including ChatGPT, Perplexity, Claude, and Gemini. The exact behaviour varies by product and query type, with some running a live search on nearly every query.
Can Live Search Agent traffic be blocked?
It can, but blocking forgoes both the revenue opportunity and a presence in AI answers. Many publishers prefer to monetise high-value Live Search retrievals while blocking low-value scrapers.
How can I tell how much Live Search Agent traffic my site gets?
You need server-side or CDN-level analytics that classify bots by type and owner. A free domain audit or analytics integration that operates at the edge can show the volume and mix, which client-side tools cannot.