How can you tell a search query came from an AI agent?

Three fingerprints, rarely combined by humans: exact-quoted fact strings (like a price with its qualifier: '"pr newswire" "$805" "400 words"'), machine-stacked search operators (one query in our data excludes ten domains with -site: operators), and verification-style permutations — the same fact queried five slightly different ways in one period, which is how agents cross-check claims.

Why do these queries get zero clicks?

Because the searcher isn't a person deciding what to read — it's an agent extracting a fact. It reads the result content, takes the figure, and synthesizes an answer for its user. The 'visit' shows up as an impression in Search Console and often as an AI-crawler fetch in server logs, but never as a click.

If there are no clicks, what's the business value?

Your figures — and your brand — become part of the AI's answer to your buyer. When an agent researching wire-service prices extracts its numbers from your comparison page, your framing and your alternative offer travel with the answer. That's citation-based visibility: influence without the click.

Is this measurable for any website?

Yes. In Google Search Console, filter queries containing quoted strings, -site: operators, or repeated exact-fact permutations, and cross-reference position vs CTR. Pair it with server-log analysis of AI user agents (GPTBot, ChatGPT-User, PerplexityBot, ClaudeBot) to see which pages agents fetch directly.

The Search Queries Only AI Agents Type — And Why 0 Clicks Is the Win

TL;DR

We pulled our Google Search Console data on July 4, 2026. Our wire-service pricing comparison now shows 407,812 impressions, and the query list behind it contains something humans don't produce: 50+ queries built like machine instructions, ranking positions 1.5–3, with zero clicks.

This is a follow-up to We Watched 108,000 AI Agents Research PR Pricing, moving from server logs to the search-results layer. Same conclusion, new vantage point: AI agents are a primary audience, and they don't click — they extract.

What Machine Queries Look Like

Three patterns from the actual query report:

1. Exact-quoted fact strings. Humans search "pr newswire pricing." Agents verify facts:

"pr newswire" "$805" "400 words"        — 188 impressions, position 2.7
"national release: $805 for 400 words"  — 25 impressions, position 6.6
"pr newswire" "membership fee" "$195"   — 13 impressions, position 3.9

Nobody types a price with its word-count qualifier in nested quotes. A retrieval system verifying a claimed figure does exactly that.

2. Stacked exclusion operators. One query in our data searches a competitor's name followed by -site:reddit.com -site:twitter.com -site:x.com … ten excluded domains in a row. That's an agent filtering UGC noise out of its source pool — a documented pattern in how AI search tools compose queries.

3. Verification permutations. The same fact, queried five ways in one window: "pr newswire" "$805", "pr newswire" "805" "400 words", "pr newswire" "national release" "$805", "national release" "$805" "pr newswire"… Humans satisfice; agents cross-check.

Across these 50+ machine-fingerprint queries we hold positions 1.5–3.4 almost uniformly — and every one of them shows 0 clicks.

Why Zero Clicks Is the Point

An agent that has your page in position 1–3 doesn't need to "click" like a browser user. It reads the indexed content or fetches the page directly (our server logs recorded ~6,300 AI-agent visits per week doing exactly that), extracts the figure, and hands its user a synthesized answer — with our comparison data, our framing, and our €49 alternative inside it.

In the impressions-era mental model, 407,812 impressions with a fraction-of-a-percent CTR looks like failure. In the Citation Economy model, it's the whole strategy working: the answer your buyer receives was assembled from your page.

Three honest caveats, because first-party data deserves them:

GSC doesn't label queries by agent vs human — the fingerprints above are strong signals, not certainties.
We can't observe which AI product issued which query; exclusion-operator syntax matches known LLM search-tool behavior, but attribution is inference.
Human zero-click behavior (answer absorbed from the snippet) also exists in this data; the query syntax is what separates the machine cohort.

How to Run This Audit on Your Own Site

GSC query report → filter for " quoted strings and -site: operators; sort by impressions. Note position vs CTR.
Server logs → count AI user agents (GPTBot, ChatGPT-User, PerplexityBot, ClaudeBot, Google-Extended) per path. Our robots.txt for AI crawlers guide covers identifying and admitting them.
Cross-reference → pages that score high on both lists are your agent magnets. Feed them: exact figures, dated claims, extractable structure.
Then track the output side — whether AI engines actually cite you. That's the closed loop Pressonify runs on every release.

What This Means for Your Content

The pages winning this traffic share three traits: exact figures (agents verify numbers, so pages with real, dated prices win verification queries), extractable structure (tables and quotable claims survive summarization), and honest comparisons (the query data shows agents researching competitors' prices — the page that answers wins the citation). It's why our per-platform guides — ChatGPT Search, Perplexity, Gemini, AI Overviews — all converge on the same content discipline, and why we built the True Annual Cost Calculator as a machine-readable fact source rather than a marketing page.

The Search Queries Only AI Agents Type — And Why 0 Clicks Is the Win