ChatGPT SEO: How the AI Determines Brand Recommendations Before Searching

2026-08-17 · Sebas K · 5 Min. Lesezeit

ChatGPT generates its own search queries with brand names before searching. These queries include brands the user didn’t mention. The choice affects which products appear. Suganthan, a developer and SEO expert, studied this for two months. Brands in the search queries are 33 times more likely to be mentioned than brands found only through search. The article explains the mechanics and offers recommendations for companies and SEO managers.

Overview

Suganthan ran an experiment. He asked ChatGPT questions about product categories like AI note-taking apps and robot vacuums. Before each search, ChatGPT generates its own query, visible in the JSON field search_queries. These queries often include brands the user never mentioned. ChatGPT decides which brands are relevant before gathering information. This shifts SEO from classic ranking to generative engine optimization.

The analysis uses 27 conversations, 13 categories, and over 3,500 pages. Four findings stand out. ChatGPT creates its own search queries that can be viewed. These queries contain brands without user input. Mention in the query increases the chance of being cited. A hard filter lets only a few pages through. The article shows how brands can appear in AI recommendations.

Prompt Analysis

The article does not break down classic user prompts. Instead, it examines the search queries ChatGPT generates. These queries serve as example prompts because they reveal the model’s search intent. An example for AI note-taking apps:

The Prompt

best AI note taking apps 2026 official pricing features Granola Notion AI Otter Fireflies Fathom Mem Limitless

Components

This search string has several functional parts:

Main term: “best AI note taking apps” – the core search query for the product category. ChatGPT reformulates the user’s question in a standardized way.
Year: “2026” – adding the current year makes results more relevant, typical SEO behavior.
Additional terms: “official pricing features” – these words filter official pricing pages and feature overviews. ChatGPT prefers manufacturer information.
Brand list: “Granola Notion AI Otter Fireflies Fathom Mem Limitless” – seven specific product names without user input. This list apparently comes from training knowledge and forms a shortlist.

ChatGPT uses these search queries for the first round of results. In follow-up searches, it specifically queries individual brands, for example with site:granola.com pricing. This runs automatically before the answer is created.

Frequently Asked Questions

Why does ChatGPT generate its own search queries?

ChatGPT uses web search for current or product-related questions to give accurate answers. It generates its own query instead of using the user’s exact words. The query includes relevant keywords and brands. This is needed because the model lacks current data and must search efficiently.

Can you view these search queries yourself?

Yes. The generated search queries are in the JSON response field search_queries (formerly search_model_queries). You can retrieve them via the browser’s DevTools – no special knowledge required. Open ChatGPT, ask a question, switch to the network tab in DevTools, and check the key.

How big is the difference between brands in the search query and those that are only found?

According to Suganthan’s data, brands in the search query are mentioned in 68.9% of the final answers. Brands that are only found via search only make it to 2.1%. That’s a factor of 33. Additionally, there are cases where brands are recommended without their website being crawled – a sign that the decision is based on internal knowledge.

How can a brand get into ChatGPT’s shortlist?

Since the shortlist apparently comes from training, brands must be publicly present: mentions in industry blogs, comparison portals, review sites, press articles, and analyst reports count. Suganthan recommends asking the category question five times and noting which brands appear each time. Brands that appear consistently have a fixed position. Otherwise, brand building and digital PR help, especially in high-quality, topic-relevant publications.

What does this mean for classic SEO optimization?

Classic SEO optimizes the website for search engines to rank organically. ChatGPT reverses the process: The brand recommendation is made before the website is accessed. A well-optimized presence only counts when the brand appears in the search query. The focus shifts from technical SEO to brand visibility and authority. Technical measures like site:domain probes are only relevant for brands on the shortlist.

Is ChatGPT’s behavior consistent across different categories?

No. The consistency varies: In established categories with clear market leaders (e.g., accounting software), the shortlist remains stable. In unclear areas (web hosting), it fluctuates strongly between sessions. The search strategy also differs: For electric SUVs, ChatGPT searches for test reports from magazines like Car and Driver or Edmunds, for software directly the manufacturer websites. The model applies different evaluation patterns depending on the domain.

How can I make my brand visible if it never appears in the search query?

Then brand building takes priority: PR, guest articles, comparison lists, product tests, reviews on independent platforms. The goal: ChatGPT takes the brand into its internal ranking as a relevant candidate. Suganthan advises redistributing the budget – technical GEO work only makes sense for brands that are already on the candidate lists.

Source

Based on this article.