For years, if you wanted to find something on the internet, you probably started with Google.
That habit is becoming less automatic.
ChatGPT has already moved beyond being just a chatbot. Its search feature can look across the web, return current information and provide links to sources. But a new development suggests OpenAI may be going further: building more of the search infrastructure behind ChatGPT itself.
Researchers at Peec AI reported in September 2026 that ChatGPT has been using an internal retrieval system called Labrador, described as a family of search indexes covering areas such as the general web, news, shopping, images, PDFs and other content types.
That does not mean OpenAI has suddenly launched a full Google replacement. OpenAI’s public documentation still describes ChatGPT Search as using multiple sources and search providers.
But if OpenAI continues developing its own web index, the implications for Google, SEO, publishers and website owners could become much bigger.
ChatGPT Started as a Chatbot. Search Changed That
When OpenAI introduced ChatGPT Search in October 2024, it positioned the feature as a way to get timely answers from the web while keeping the conversational experience of ChatGPT.
OpenAI said at launch that ChatGPT Search used third-party search providers alongside content supplied directly by partners. The idea was not simply to show a traditional list of links. ChatGPT could search based on the question, combine information and present an answer with relevant sources.
That already changed the relationship between a user and a website.
Instead of searching Google, opening five pages and comparing them, a user could ask ChatGPT the question and receive a synthesized answer with links to the underlying sources.
OpenAI’s original announcement of ChatGPT Search explains how that system was introduced.
Since then, ChatGPT Search has become another discovery channel that publishers have to consider.
I’ve already looked at this from the publisher’s side in Can You See When ChatGPT Cites Your Website? What You Can Actually Track.
But the Labrador development takes the question one step further.
What Is Labrador?
According to research published by Peec AI, Labrador is an internal search index used by ChatGPT rather than simply an interface sitting on top of another search engine.
The research describes multiple specialized indexes, including general web pages, news, shopping, images, YouTube, PDFs, local information and other categories.
That is important because a search index is fundamentally different from simply sending every query to another search engine.
A company operating its own index has more control over how pages are discovered, stored, retrieved and presented.
Peec AI also reported seeing evidence that ChatGPT was testing its own index against other retrieval sources, particularly for shopping-related searches.
However, there is an important distinction here.
Labrador is reported by researchers; OpenAI has not publicly published a complete technical specification for it.
What OpenAI has publicly confirmed is that it operates OAI-SearchBot, a crawler specifically used to surface websites in ChatGPT Search.
OpenAI’s current crawler documentation says that websites which block OAI-SearchBot will not be shown in ChatGPT search answers, although they can still appear as navigational links.
OpenAI’s crawler documentation provides the current technical details for publishers.
Why Would OpenAI Build Its Own Search Index?
There are several practical reasons.
The first is control.
If an AI assistant depends entirely on another company’s search index, it has less control over what information is available, how quickly it is updated and how results are retrieved.
An internal index gives OpenAI another layer of control over that process.
The second is speed and scale.
ChatGPT does not behave exactly like a traditional search engine. People ask long, conversational questions, follow up with additional context and expect the system to understand what they mean rather than simply match keywords.
An AI-focused retrieval system can therefore be designed around those kinds of queries.
The third is specialization.
Shopping searches, news searches, local searches and general web searches do not necessarily need to work in exactly the same way.
A system with specialized indexes can potentially handle different types of information differently.
That matters because ChatGPT is increasingly being used for product research and other activities that traditionally happened through search engines.
This Is Where Things Get Interesting for Google
It would be premature to say ChatGPT is replacing Google.
Google still operates one of the world’s largest search infrastructures, and millions of websites continue to depend heavily on Google Search for traffic.
The more interesting change is that Google may no longer be the only major gateway between people and the web.
Imagine someone searching for a product comparison.
The traditional journey might look like this:
- Search Google.
- Open several websites.
- Compare reviews.
- Check prices.
- Make a decision.
An AI search journey can look very different:
- Ask a detailed question.
- Receive a synthesized answer.
- Inspect the sources cited by the AI.
- Continue asking follow-up questions.
- Visit only the websites that appear relevant.
The website has not necessarily disappeared.
But the intermediary has changed.
SEO May Become More Complicated
This is probably the part website owners should pay the most attention to.
For years, SEO strategy was largely built around one major question:
How do I get my website to rank in Google?
That question still matters.
But it is increasingly being joined by another:
How does my website become a useful source when an AI system answers a question?
Those are not necessarily the same thing.
A page can perform well in traditional search and still not become a frequently cited source in an AI answer. Likewise, an AI system may retrieve information from sources that do not appear at the top of a conventional search results page.
That is one reason I’ve argued in How to Find Video Content Gaps in AI Search and Get Cited that publishers should test actual AI search behaviour instead of assuming that traditional SEO rankings tell the entire story.
The search ecosystem is becoming more fragmented.
You can now have visibility across Google, Bing, ChatGPT, Perplexity, Gemini and other AI-driven discovery systems, and they do not necessarily retrieve or present information in the same way.
What Happens to Websites If AI Becomes the Middleman?
This is where the biggest concern for publishers comes in.
A website traditionally gets value from being discovered and visited.
Someone searches for something, sees the website, clicks through and reads the article.
AI search can put an additional layer between the reader and the publisher.
The AI may use information from several websites, summarize it and give the user an answer before the user ever visits those sites.
That does not make websites irrelevant. In fact, AI systems still need useful sources to retrieve information from.
But it can change how much traffic a source receives for the information it contributes.
For publishers, that means the value of being cited, mentioned or linked by an AI system could become increasingly important.
It also means that simply publishing another generic article on a topic may not be enough.
OpenAI Is Already Giving Website Owners a Technical Signal
This part is not speculation.
OpenAI has publicly documented OAI-SearchBot, its crawler for ChatGPT Search.
OpenAI says publishers who want their websites to be eligible for ChatGPT Search should allow OAI-SearchBot to crawl their content. It also says that referral traffic from ChatGPT can be tracked in analytics because ChatGPT adds utm_source=chatgpt.com to referral URLs.
That creates something website owners can actually monitor today.
Check whether ChatGPT is sending visitors to your website.
Check which pages receive that traffic.
Test important questions in ChatGPT Search and see whether your pages appear among the sources.
And check your site’s crawler settings before assuming AI search can access everything.
OpenAI’s publisher FAQ explains how website discovery and referral tracking currently work.
What Website Owners Should Do Now
I would not recommend abandoning traditional SEO because of Labrador.
That would be an overreaction.
Instead, website owners should start treating AI discovery as another part of their overall search strategy.
Keep your Google SEO foundation strong
Google Search is still important. Technical SEO, useful content, internal linking, page experience and legitimate authority still matter.
Make your content genuinely useful
AI systems have an enormous amount of generic information available to them. Publishing another article that says the same thing as everyone else gives your site little reason to become a distinctive source.
Original testing, first-hand experience, examples, data, comparisons and clear explanations give your content more substance.
Check your AI search visibility
Search for the questions your audience actually asks.
Don’t only search your brand name.
Test informational questions, comparisons and problem-solving queries, then record which sources ChatGPT cites.
Don’t block OAI-SearchBot by accident
If you want your website to be eligible for ChatGPT Search, make sure your robots.txt and other security layers aren’t preventing OAI-SearchBot from accessing your content.
Allowing the crawler does not guarantee a citation or traffic. It simply removes one technical barrier to being discovered through that route.
Watch your analytics
AI referral traffic may look small compared with Google traffic today, but it is worth measuring rather than guessing.
Over time, you may discover that certain articles attract visitors from AI systems even when their traditional search performance looks very different.
Don’t Start Writing for AI Instead of People
There is a temptation whenever a new search system appears to start creating content specifically designed to manipulate it.
That is the wrong lesson to take from this development.
If AI systems are becoming another way people discover information, the long-term advantage is still likely to come from publishing information that is genuinely worth discovering.
That means clear answers, original information, accurate details and useful experiences.
It also means building a site with enough depth that readers have a reason to visit more than one page.
Internal linking becomes particularly important here.
For example, a site discussing AI search can naturally connect an article about AI citations with another about finding content gaps, rather than treating every article as an isolated page.
That creates a stronger information network for both readers and search systems.
The Web Could End Up With More Than One Search Gateway
The most important part of the Labrador story is not whether OpenAI eventually launches a product called a “search engine.”
It is that AI companies are increasingly building their own infrastructure for discovering and retrieving information from the web.
ChatGPT already has web search. OpenAI already operates a search crawler. Researchers report that OpenAI has developed its own internal retrieval index.
Those pieces point toward a broader shift in how information may be discovered online.
Google is still Google. Traditional SEO is not suddenly obsolete.
But website owners should probably stop thinking of search as one doorway.
The web may be moving toward a world where people discover the same website through several different systems, each with its own way of deciding what information deserves to be surfaced.
For publishers, that makes the goal bigger than simply ranking.
The goal is to build a website that remains useful wherever people go to find answers.
And if OpenAI continues building out its own search infrastructure, that destination could increasingly include ChatGPT.
Conversation (0)