I think I’m on to something. Something strange is happening in Search Console and Bing Webmaster Tools.

I could be wrong, but the patterns are clear. 

SEOs I need you to go into Search Console and either prove me wrong or back me up. 

It started with a few search queries that didn’t feel human. 

They were too long. Too structured. Too formal. 

Almost like someone had pasted an assignment or research prompt directly into the search bar.

But what if that’s exactly what happened?

Is it a person? Or is it AI? Is it an AI assistant? Is it AI Mode, is it Gemini? 

Or is it a damn AI rank / response tracker?

These kinds of queries are starting to appear more frequently in performance reports.

They often register impressions without clicks, and they read like machine-generated instructions.

My working theory is that they could be signs of AI-led search. 

Tools that use search engines not for browsing, but for extracting information.

If that’s the case, it has big implications for how we think about SEO, search behavior, and content visibility.

Let me show you what I found.

What I Found in Google Search Console

The first clue came from a query that showed up in my GSC performance report:

“evaluate the financial planning and analysis software company [name removed] on corporate performance management”

At a glance, it looks like someone is doing detailed research.

But the phrasing is too deliberate and structured to be organic. 

This reads like a prompt. And not just any prompt: one generated by a tool working through a checklist.

And that’s NOT the only query. 

There are loads of similar queries. 

Here’s another one.
“evaluate the financial planning and analysis software company [Hidden] on financial modeling platforms”

Here’s a bunch more:

That’s when the idea clicked: what if this wasn’t a person searching, but a tool?

An agent running a retrieval task.

 A research assistant powered by an LLM. 

A system trying to build a report, summary, or recommendation by scraping what’s available online, and doing it through Google Search.

By filtering queries with keywords like “evaluate,” “assess,” or “analyze,” I started noticing several impressions tied to long, formal queries with little to no clicks.

That’s not normal for typical user behavior,  but it’s completely expected if the searcher is a machine.

Then It Started Showing Up in Bing

After spotting the patterns in Google, I checked Bing Webmaster Tools, and sure enough, the same thing was happening there.

One query in particular stood out:

“create a report (no more than two pages) to the CFO that outlines which HIM functions could be managed from home and which functions will need to stay located within the [Private] office. research the best practices associated with the functions that are selected to be managed from home and explain how to incorporate these practices into managing a remote workforce. evaluate whether or not [Private] management staff can work remotely as well and if so, outline the portion of time each manager may work at home. provide appropriate references for the researched articles.”

That’s not a search. That’s a task brief.

Possibly copy-pasted from a course assignment or internal workflow. 

Or more likely: generated and submitted by an AI system trying to complete it.

There’s more.

“the evaluation should include specific scenarios in which the approach is expected to perform continuously worse to check its performance even in less than ideal / worst case scenarios.”

I’m not done yet. Here’s another.

These are not typical searches.

Just like with Google: these showed impressions without clicks.

They weren’t looking to engage. 

They were there to extract.

Why I Think These Are Traces of Deep Research Tools

These aren’t just outliers. They’re signals.

The phrasing, structure, and zero-click pattern all point to something bigger: automated tools using search engines as research assistants.

Here’s why I don’t think these are traditional human search queries:

A) They’re too long. People don’t type out multi-sentence instructions in a search bar.

B) They’re task-based. The language mirrors prompts — not questions.

C) They follow a structured format. Often tied to corporate, academic, or technical workflows.D) They trigger impressions, not clicks. Which suggests someone (or something) is retrieving — not browsing.

The user never sees your page. But the system does, long enough to reference or summarize it.

That’s why the impression shows up.

And this isn’t just theory. 

It’s consistent with how modern retrieval-augmented generation (RAG) systems work. 

They use tools like Bing or Google as input channels, grab context, and pass it back into the model to complete a task, whether it’s writing a report, summarizing best practices, or assembling answers.

We may be witnessing the early signs of how agentic workflows quietly reshape the search landscape.