People aren’t just Googling anymore; they’re asking ChatGPT, Claude, and Perplexity to make decisions for them. “What’s the best CRM for startups?” “Which landing page builder should I use?”
The challenge is that you can’t see where ChatGPT looks before it picks a winner: the queries it uses to retrieve information and build AI recommendations are completely invisible.
Until now.
A developer named Zack Notes built a Chrome extension that reveals the exact ChatGPT search queries running behind the scenes.
This, in essence, should fundamentally change how you think about AI content strategy. Because when you see what ChatGPT actually searches for, you realize most brands are showing up in all the wrong places.
TL;DR
- ChatGPT can turn a single user prompt into several targeted web searches, often looking for comparisons, pricing, features, and reviews before making a recommendation.
- Ahrefs’ 2026 analysis of 66.6 million ChatGPT responses found that 14.4% contained query fan-outs, with approximately 1.6 search queries per response when fan-out occurred.
- Transactional content serves two purposes: capturing high-intent Google searches and providing information ChatGPT can retrieve when comparing products.
- CXL adapted an existing Chrome extension to capture ChatGPT’s search queries and turn them into content suggestions. Those queries can help identify missing comparison pages, product information, and third-party coverage.
Table of contents
- How does ChatGPT turn one question into multiple search queries?
- Why does transactional content matter for AI recommendations?
- Why does third-party validation matter for ChatGPT recommendations?
- How to build a Chrome extension that reveals ChatGPT search queries
- How can marketers use ChatGPT’s hidden search queries for content planning?
- What makes a brand more likely to get recommended by ChatGPT?
How does ChatGPT turn one question into multiple search queries?
ChatGPT can break a single question into multiple targeted searches, a process known as query fan-out. Instead of searching only what the user typed, it can look for comparisons, pricing, features, and other information needed to build an answer.
Ask ChatGPT, “What is the best SEO keyword research tool?” and it doesn’t just search that phrase. It runs:
- “Best SEO keyword research tools 2025”
- “SEO keyword research tool comparison features pricing”
Ask “HubSpot vs Salesforce for startups” and it searches:
- “HubSpot vs Salesforce for startups comparison CRM HubSpot Salesforce benefits for startups”
- “HubSpot & Salesforce features pricing scalability startups”
The original experiment showed us what this looked like. Ahrefs’ October 2026 study gives us a much bigger picture. Analyzing 66.6 million responses and 15 million fan-out queries from August and September 2026, Ahrefs found that 14.4% of responses contained fan-out queries. Those responses generated an average of 1.6 queries each. The searches also showed a growing emphasis on official sources, pricing, and reviews.
ChatGPT is validating its recommendation before it gives one.
It’s looking for comparison data, pricing information, feature breakdowns, and proof points across multiple search angles. In other words, if your brand isn’t showing up in these validation searches, you’re not getting recommended. Period.
The AI SEO extension works by intercepting the Fetch API calls between ChatGPT’s frontend and OpenAI’s backend. Think of it as a radio interceptor: you see the final broadcast on screen, but the actual search queries are buried in the raw signal.
The extension ‘listens’ to the data streaming into your browser and scans it for a specific tag called “search_model_queries”. Whenever it spots that tag, it grabs the search terms, cleans them up, and shows them to you in a separate box.
Why does transactional content matter for AI recommendations?
Transactional pages help brands capture high-intent Google searches while giving ChatGPT the information it needs to compare products. Pricing guides, competitor comparisons, feature breakdowns, and case studies can serve both search journeys.
This invisible validation process connects directly to something we’ve been seeing in the data: transactional content protects you from two threats at once.
Ahrefs data shows that AI Overviews (the main reason behind declining informational traffic) appear far less frequently on transactional queries.
When someone searches “Salesforce pricing” or “HubSpot vs Pipedrive,” they still click through to actual pages. Google knows AI can’t close that deal.
But here’s the part most marketers miss: those same transactional pages are exactly what ChatGPT searches for when building recommendations.
- Comparison pages;
- Pricing guides;
- Feature breakdowns;
- Third-party reviews;
- Case studies with specific results.
These aren’t just SEO assets anymore; they’re LLM validation sources.
Build them right, and you’re defending against traffic loss from AI Overviews while simultaneously positioning your brand to win AI recommendations.
The inverse is also true.
If you’ve spent years building thin informational content that ranks but doesn’t convert, you’re exposed on both fronts. AI Overviews are eating that traffic, and ChatGPT isn’t finding the proof points it needs to recommend your brand.
Why does third-party validation matter for ChatGPT recommendations?
Third-party sources give ChatGPT another way to find and check information about a brand. Reviews, independent comparisons, industry publications, and credible mentions can help establish whether a product is relevant to the recommendation being made.
The AI SEO extension reveals something else: ChatGPT actively hunts for external validation.
When it’s deciding what to recommend, it doesn’t just look at your website. It searches for third-party mentions, reviews, comparisons published by others, and credible sources discussing your product. This is why Digital PR matters more than it ever did for traditional SEO.
AI doesn’t care about your brand story or your founder’s vision. It cares about proof. Can it find your name in Software Advice reviews? Are you mentioned in G2 comparisons? Did TechCrunch or an industry blog cover your latest feature release?
If the answer is no, you’re making it harder for ChatGPT to find evidence that supports recommending your product. Your website can tell the story, but independent sources give that story something to stand on.
This changes the ROI calculation on PR entirely.
You’re not just earning backlinks or domain authority. You’re creating the external signals that influence whether an AI will ever suggest your product to a potential customer.
How to build a Chrome extension that reveals ChatGPT search queries
We took Zack’s original Chrome extension and used ChatGPT to add a feature that turns the captured searches into content suggestions. The build keeps the original query-capture functionality, then uses the extracted searches to suggest content pillars and topics. Here’s how we put it together.
While testing Zack’s extension, we realized it could go further. Seeing the hidden queries is valuable, but what if it could also suggest content pillars and topics based on those queries?
We built that version in 30 minutes. Here’s exactly how.
Step 1: Upload and prompt ChatGPT.
We gave ChatGPT the original extension file and asked it to recreate the functionality while adding an AI-powered feature that generates content suggestions from the extracted queries. Since this requires an API call, we instructed it to integrate our OpenAI API key in the backend.
You can get Zack’s original extension files on GitHub. Download the repository and give ChatGPT the source files. The important part is keeping the existing query interceptor intact while adding a feature that turns the extracted searches into content suggestions.
Step 2: Download and unzip the generated file.
ChatGPT outputs a ZIP file. Extract it locally. Make sure the extracted folder contains a manifest.json file. Chrome needs this to recognize the extension when you load it in the next step.
Step 3: Load the extension in Chrome.
Go to chrome://extensions, enable Developer mode, click “Load unpacked,” and upload the folder. If you hit an error, copy the message back into ChatGPT and ask it to fix the code. Then repeat the process and upload the updated folder.
Step 4: Generate your OpenAI API key.
Visit platform.openai.com/api-keys, create a new key, and confirm billing is configured so API calls don’t fail.
Step 5: Configure the extension.
Open it from your Chrome toolbar and paste in your API key.
One thing we’d change for a production version: API-key handling. In this prototype, we entered the OpenAI API key directly into the extension. That was how we tested it, but it’s not a secure setup for something you’re going to distribute or use long-term.
For a reusable build, have ChatGPT route the content-generation requests through a private backend endpoint, with the API key stored as an environment variable. The extension sends the extracted queries to that endpoint, and the backend handles the OpenAI API call. That way, the key stays out of the browser.
This preserves the original experiment and screenshot while adding a necessary technical update. OpenAI’s API-key security guidance specifically recommends routing browser-based applications through a backend.
Step 6: Test it live.
When we asked, “What is the best landing page builder?” ChatGPT ran searches for “best landing page builders 2025” and “landing page builder comparison features pricing.” The extension then generated content pillar suggestions based on those exact queries.
To test your own version, open ChatGPT with the extension enabled and ask a product recommendation question that triggers web search. Check that the extension captures the search queries before it generates content suggestions. Then try a comparison question to see whether the suggested topics change with the searches.
If nothing appears, first check whether ChatGPT actually used web search. Reload the extension and refresh ChatGPT before testing again. Because the interceptor depends on ChatGPT’s response structure, the code may need updating when that structure changes.
The experimental build has some bugs, but it works. You can follow these steps yourself or download our version to test.
We used a similar approach to build an AI Overview gap analysis extension that compares what Google surfaces against the information on your own page.
How can marketers use ChatGPT’s hidden search queries for content planning?
Start by looking at the queries ChatGPT runs when buyers ask about your category. They can reveal the comparisons, pricing information, reviews, and product details that matter during AI-assisted research. Then check where your brand already appears and where important information is missing.
ChatGPT is already making recommendations. It’s already deciding which brands get surfaced and which ones don’t.
With this AI SEO extension, you can see exactly what ChatGPT searches for when someone asks about your category.
Here’s what to do next:
- Run queries in your category and map the search patterns. Ask ChatGPT the questions your customers actually ask. “Best [your product category] for [use case].” “Product A vs Product B.” Watch what it searches for in the background. Those are your gaps.
- Audit your transactional content against the query patterns. Do you have real comparison pages that pit you against competitors? Not fluffy “vs” pages with marketing copy: actual feature-by-feature breakdowns with pricing. Do you have detailed case studies with specific results? ChatGPT is looking for proof, not promises.
- Invest in external validation sources. Get your product listed and reviewed on G2, Capterra, and Software Advice. Pursue mentions in industry publications. Earn those third-party signals. An AI won’t recommend what it can’t validate.
- Build the content ChatGPT can’t find about you. If the hidden queries reveal it’s searching “your-brand pricing” or “your-brand review” and those pages don’t exist or are thin, you’re losing recommendations while reading this post.
This isn’t theoretical. Every time someone asks ChatGPT for a recommendation in your space, and you’re not in the results, a competitor is gaining that consideration.
What makes a brand more likely to get recommended by ChatGPT?
Brands are easier to recommend when ChatGPT can find useful information to back them up. Clear pricing, genuine product comparisons, customer reviews, and credible third-party coverage give it something to work with. Keyword density and another hundred generic blog posts won’t solve that problem.
Although some people still Google, an increasing number are asking AI for recommendations. And AI doesn’t care about your keyword density or meta descriptions. It cares about whether it can validate you through multiple searches across comparison pages, reviews, pricing data, and third-party mentions.
The shift from visibility to credibility isn’t coming. It’s here.
The brands that succeed aren’t the ones with the most content. They’re the ones that show up in the right places when AI is doing its due diligence. The ones that built comparison pages that actually compare, pricing pages that inform, and external validation that really exists.
Building the content infrastructure that wins AI recommendations isn’t a 90-day sprint. It’s a multi-year investment in becoming the most credible, validated, provable option in your category.
Start by wiretapping the recommendation process. Then join the conversation your competitors are already a part of.
Frequently asked questions
What is ChatGPT query fan-out?
Query fan-out is when ChatGPT turns a user’s question into one or more targeted web searches to gather information for its answer. A request for the best CRM might trigger searches about pricing, features, competitors, and specific use cases. These searches can reveal what information ChatGPT is looking for when evaluating products.
Can you see the exact search queries ChatGPT uses?
You can inspect search queries exposed in ChatGPT’s web-search response data using tools such as the Chrome extension featured in this article. It extracts the search_model_queries field and displays the captured searches. The tool only shows queries available through that mechanism, and its functionality depends on ChatGPT’s current response structure.
How can hidden ChatGPT queries improve your content strategy?
They help reveal the questions and comparisons ChatGPT searches when researching products in your category. Instead of relying entirely on traditional keyword lists, you can identify missing pricing information, competitor comparisons, feature breakdowns, or third-party coverage. Use those patterns to prioritize content that answers real buyer questions.
Does ChatGPT need third-party reviews to recommend a product?
Not necessarily, but independent reviews and comparisons can give ChatGPT additional information when evaluating products. Sources such as G2, Software Advice, industry publications, and credible reviews help buyers verify claims beyond a company’s own website. Accurate product information on your own website remains important too.
What to do next
If you want to turn these hidden queries into a content strategy, CXL’s Content strategy for LLM visibility and changing search habits course goes deeper into reverse-engineering AI citations, finding content gaps, and building visibility beyond your own website. For the comparison content ChatGPT frequently searches for, Optimize pages for AI search with GEO/AEO covers how to audit your presence and build useful product comparisons that buyers and AI search tools can understand.
