People don't just Google things anymore. They ask ChatGPT. They ask Claude. They ask Perplexity. And they let these tools make decisions for them.

"What's the best CRM for a startup?" "Which landing page builder should I use?"

You've probably typed something like this yourself. And whatever answer you got back, you probably trusted it. That's the interesting part. We trust these answers, but we have no idea how they're made.

Here's the real problem. You can't see what ChatGPT looks at before it picks a winner. The searches it runs behind the scenes, the ones that decide which brand gets named and which one gets left out, are completely hidden from you. You only ever see the final answer. You never see the homework. Until now.

A developer built a small Chrome extension that pulls back the curtain. It shows you the exact search queries ChatGPT runs behind the scenes, before it gives you a recommendation. 

And once you see those hidden queries, one thing becomes obvious very fast: most brands are putting their effort in the wrong places. They're still writing content for the old version of search. 

Not the version that's happening right now.

Here, we will walk you through what those hidden queries teach us. We'll go subtopic by subtopic, in plain language, so you walk away with a clear picture of what to actually do next.

ChatGPT's Hidden Search Queries

ChatGPT's Hidden Search Queries are when it rewrites your prompt into multiple background queries known as fan-out queries before generating an answer. 

What Are Fan-Out Queries?

ChatGPT rewrites your original prompt into multiple separate search angles (spanning web, shopping, or images) to gather comprehensive context before generating an answer.


Studies show that background sub-queries occur in roughly half of all searches, and these generated queries are typically 31% longer and phrased differently than the user's initial input.

If your content is only optimized for the exact short-tail phrase a user types rather than the expanded semantic fan-out queries, AI search engines may overlook your website during source selection.

How to Reveal ChatGPT's Hidden Search Queries

 You can unearth these backend queries using developer tools or specialized browser extensions:
1. Using Browser Developer Tools: Open a new chat in ChatGPT, enter your prompt, and let the response finish.
2. Copy the unique conversation ID from the URL (the alphanumeric string appearing after /c/).
3. Right-click the page, select Inspect, and navigate to the Network tab.
4. Filter by Fetch/XHR, paste your conversation ID or look for the conversation JSON request, and open the Response tab to inspect the parsed search payloads and background tool calls.

Most people assume that when you ask ChatGPT a question, it runs one search, reads a bit, and answers you. That's not what happens. One simple question from you can trigger several separate searches happening quietly in the background, before you ever see a reply.

Here's an easy example. Say you type: "What is the best SEO keyword research tool?"

ChatGPT doesn't just search that one phrase. It actually runs more than one search, such as:

  • "best SEO keyword research tools 2025"
  • "SEO keyword research tool comparison features pricing"

Now try a different kind of question. Say you ask: "HubSpot vs Salesforce for startups."

Behind the scenes, ChatGPT might search something like:

  • "HubSpot vs Salesforce for startups comparison"
  • "HubSpot and Salesforce features pricing scalability for startups"

Look closely at those queries. Notice what they have in common. They're not casual questions. They're research questions. ChatGPT is a comparison. It's checking prices. It's looking for features side by side. It's gathering proof.

ChatGPT is checking its own homework before it gives you an answer. It doesn't just guess who's "the best." It goes out, gathers comparison data, pricing details, and feature breakdowns from multiple angles, and only after all that does it settle on a recommendation.

This one fact changes everything about how you should think about content. If your brand doesn't show up somewhere in those background searches, you're not just ranked lower. You're not in the running at all. It doesn't matter how great your product actually is. 

If ChatGPT can't find proof of it during its research step, it simply picks a brand it can find proof of instead. That's not personal. That's just how the process works.

So how does anyone actually know this is happening? This is where the extension comes in.

The tool works by watching a specific part of the technical conversation happening between your browser and OpenAI's servers. Every time you chat with ChatGPT, data is constantly flowing back and forth behind the scenes, way more data than the words you actually see on your screen. 

Buried inside that flow is a small, specific tag called "search_model_queries." That tag holds the actual search terms ChatGPT used before writing its answer to you.

Why Transactional Content is Now a Two-Birds-One-Stone Play

Now let's talk about a type of content you might already have, but probably haven't given enough credit to: transactional content.

What counts as transactional content? It's the type of page someone looks at right before they're about to buy something, or right before they're about to pick between two options. 

Think of comparison pages. Pricing pages. Feature breakdown pages. Detailed case studies with real numbers. Reviews written by someone other than you.

This kind of content is suddenly doing double duty, and here's why.

Google's AI Answers Tend to Skip These Pages

Data from Ahrefs shows something interesting. Google's AI Overviews, the short AI-written summaries that now appear above regular search results, show up far less often on transactional searches than on plain informational ones. 

When someone searches something broad, like "what is a CRM," Google is happy to just answer that directly in an AI box, and the person never clicks anywhere. 

But when someone searches something transactional, like "Salesforce pricing" or "HubSpot vs Pipedrive," Google tends to still send that person to a real page. It makes sense if you think about it. An AI summary can explain what a CRM is. It can't actually close a sale. A real page can.

This is Exactly The Content ChatGPT Searches for When Deciding What to Recommend. 

Go back to the hidden queries we just talked about. Look at what they actually are. Comparison searches. Pricing searches. Feature searches. That's transactional content, through and through. So the very same pages that protect your traffic from Google's AI Overviews are also the pages that win you a recommendation from ChatGPT.

That's a rare situation. Usually you have to choose which problem to solve. Here, one type of content solves two problems at once. Build strong comparison pages, honest pricing pages, and detailed case studies, and you're covered on both fronts.

Now let's flip this around, because the opposite is also true, and it's worth sitting with for a second.

If you've spent years pumping out thin, generic blog posts, ones that maybe rank okay but never really say anything specific, you're now exposed on two sides at once. Google's AI Overviews are quietly eating the traffic those posts used to bring in. And at the same time, ChatGPT can't find any proof points inside your content to recommend you with. You lose the click, and you lose the recommendation. Both at the same time.

A simple way to check where you stand: open a private browser window right now. Search your own brand name plus the word "pricing." Then search it again with "review 2025" added. What comes up? If you see thin, outdated, or missing pages, that's exactly the kind of gap both Google's AI Overviews and ChatGPT will notice and exploit in favor of a competitor.

Digital PR Just Became Table Stakes

Here's a piece that catches a lot of marketers off guard.

ChatGPT doesn't only look at your own website when it's deciding what to recommend. It actively goes looking for what other people are saying about you, out on the wider web.

Why would it do that? Think about it from ChatGPT's side. Anyone can write "we're the best" on their own homepage. That claim, on its own, doesn't prove anything. So instead of just trusting your marketing copy, ChatGPT goes searching for outside confirmation. It looks for third-party mentions. It looks for reviews written by real customers. It looks for comparisons published by people who have no reason to favor you. It looks for credible outlets, like TechCrunch or a respected industry blog, actually covering your product.

This is exactly why digital PR now matters more than it ever did for traditional SEO.

Picture it as a checklist ChatGPT is quietly running through. Can it find your name inside a Software Advice review? Are you mentioned in a G2 comparison article? Did a real publication write about your latest product update?

If the honest answer to all of that is "no," you're essentially invisible to the recommendation layer of AI search. And here's the important part: ChatGPT won't risk its own credibility by recommending something it can't back up. If it can't find you validated through more than one outside source, it will simply choose a competitor it can verify instead, even if that competitor isn't actually better than you.

This changes how you should calculate the return on your PR efforts. It's easy to think of PR as just a way to earn backlinks, or to look good, or to build "domain authority" as an abstract SEO number. But that undersells what's actually happening now. Every mention you earn is a small signal that helps decide whether an AI system will ever suggest your product to a future customer. PR isn't just brand-building anymore. It's becoming part of your AI visibility infrastructure.

A few practical starting points, if you're building this proof layer from scratch:

  • Claim your profile and encourage honest reviews on the review sites people in your industry actually check.
  • Reach out to writers and blogs in your space with something genuinely newsworthy, not just a generic press release nobody wants to read.
  • Watch for comparison articles other people write about your category, and make sure your brand is represented fairly and accurately in them.
  • Every few months, simply search your own brand name and see what shows up. If it's mostly your own website, that's your signal to invest here.

Building Your Own Query Interceptor

Seeing these hidden queries is genuinely useful on its own. But it raises an obvious next question: what if the tool could go a step further and actually suggest content ideas based on those queries?

That upgraded version doesn't take long to build. 

Here's exactly how it was done, broken into six simple steps.

Step 1: Upload and Prompt ChatGPT

Start by giving ChatGPT the original extension's code. Ask it to recreate the same functionality, but with one addition: a feature that automatically suggests content ideas based on the search queries it just extracted. Since generating those suggestions requires an API call, tell ChatGPT to wire in your own OpenAI API key on the backend.

Step 2: Download and Unzip the Generated File

ChatGPT will hand you back a ZIP file containing all the code it wrote. Save it somewhere you'll remember, then unzip it so the folder is ready to load.

Step 3: Load the Extension in Chrome

Here are simple steps to load the extension in Chrome:

  • Open a new browser tab and go to chrome://extensions. 
  • Turn on "Developer mode," usually a toggle in the top corner of the page. 
  • Click "Load unpacked," then select the unzipped folder.
  • If Chrome throws an error at this point, don't panic. 
  • Copy the exact error message, paste it back into ChatGPT, and ask it to fix the code. 
  • Then repeat the load step with the updated folder. This back-and-forth is completely normal.

Step 4: Generate your OpenAI API Key

Go to platform.openai.com/api-keys and create a new key. While you're there, double-check that billing is properly set up on your account. If billing isn't configured, the extension's API calls will simply fail, often without a clear error message telling you why.

Step 5: Configure the Extension

Click the extension's icon in your Chrome toolbar. Paste in the API key you just created. That's the whole step.

Step 6: Test It Live

Now try it out for real. Ask ChatGPT something simple, like "what is the best landing page builder?" Watch the extension's panel as ChatGPT works. You should see the background searches appear, things like "best landing page builders 2025" or "landing page builder comparison features pricing." If you added the content-suggestion feature in step 1, this is where it'll turn those exact queries into content ideas for you.

Don't expect a perfectly polished tool. This kind of experimental build usually has a few rough edges and small bugs. That's fine, and it's not really the point. The real value is seeing, with your own eyes, exactly what an AI tool searches for in your category. Once you've seen it, you can't unsee it, and you can start building content around what actually gets found.

What This Means for Your AI Content Strategy

ChatGPT is already making recommendations, right now, today. It's already deciding which brands get mentioned by name and which ones get quietly skipped. This isn't a future trend you need to prepare for. It's already happening, every time someone asks it a "what's the best" question in your category.

With the kind of extension we just walked through, you can see exactly what ChatGPT searches for when someone asks about your space. That visibility is genuinely rare, and it's worth acting on. 

Here's what to actually do with it:

1.Run Queries in Your Own Category, And Map the Patterns You See: Ask ChatGPT the same kinds of questions your real customers ask. Pay close attention to what it searches for in the background. Whatever it can't find well for your brand, that's your gap, and it's now a clear, specific one instead of a vague feeling.

2. Audit Your Transactional Content Against Those Exact Patterns: Be honest with yourself here. Do you have a real comparison page that stacks you up fairly against your competitors? Not a fluffy "why we're better" page full of marketing language, but an actual feature-by-feature breakdown, with real pricing. Do you have case studies with specific, believable numbers in them? Remember, ChatGPT is looking for proof. Not promises.

3. Invest in External Validation, Deliberately: Get your product properly listed and reviewed on the platforms people in your industry actually check, like G2, Capterra, or Software Advice. Pursue genuine mentions in industry publications. Earn those third-party signals on purpose, rather than hoping they happen by accident. An AI system won't recommend what it can't validate from more than one place.

4. Build the Content ChatGPT Currently Can't Find About You: If you run the hidden queries and notice it's searching for something like "[your brand] pricing" or "[your brand] review 2025," and that page either doesn't exist or is thin and outdated, that's not a small thing. It's an active, ongoing loss. Somewhere out there, right now, someone is asking ChatGPT for a recommendation in your space, and a competitor is winning that moment instead of you.

This isn't a theoretical exercise. It's happening in real time. Every day you wait to fix these gaps is another day a competitor quietly becomes the "default answer" in your category instead of you.

AI Doesn't Recommend the Best Brands. It Recommends The Validated Ones

Let's end with the one idea worth remembering above everything else in this guide.

Plenty of people still use Google the old way. But more and more people are simply asking AI tools for a recommendation and trusting the answer. And here's the thing that's hard for a lot of marketers to accept: AI doesn't care about your keyword density. It doesn't care how polished your meta description is. It cares about one thing only: whether it can actually validate you. It checks this through multiple searches: comparison pages, honest reviews, real pricing data, and mentions from people who aren't you.

This shift, from simply being visible to actually being credible, isn't something coming down the road. It's already here, happening in the background every single day.

The brands that come out ahead in this new environment aren't the ones publishing the most content. They're the ones that show up in the right places when an AI system is quietly doing its due diligence. They're the ones with comparison pages that genuinely compare things, pricing pages that actually inform instead of hide the real numbers, and outside validation that's real, not manufactured.

One honest note before you go: building this kind of content and proof, the kind that wins AI recommendations, isn't something you finish in a 90-day sprint. It's closer to a multi-year habit. It's the slow, steady work of becoming the most credible, most provable, most validated option in your category. But every brand in your space is really only just starting this race right now. That's actually good news for you.

Start simple. Go run some queries yourself. Watch what ChatGPT actually searches for in your category. Then go join the conversation your competitors are, whether they realize it yet or not, already quietly a part of.

Related Searches: