A weekend test found every link in Google AI Mode flight answers pointed to Google Flights. Search Engine Roundtable's Barry Schwartz tested Google AI Mode for flight information over the weekend and found 100 percent of the links in the responses went to Google Flights. The result raises fresh self-preferencing questions for Google.
The test was simple and the result was stark. Schwartz asked Google AI Mode for flight information and checked where the links in the answers went. Every one of them pointed to Google Flights. Not most of them, and not the top few with a mix below. All of them. For a product that presents itself as an AI assistant drawing on the web, a 100 percent share for a Google-owned property is hard to ignore.
It is worth being precise about what this test does and does not prove. It was one tester's weekend experiment with flight queries, not a large-scale study across many query types and many days. AI Mode answers can vary by query, by user, and by moment. But the pattern it found is exactly the pattern regulators and rivals have warned about: when Google answers a question in its own AI product, the links lead back to Google.
That is why a small weekend test can carry outsized weight. It gives a concrete example to a debate that is usually argued in abstractions. Instead of discussing self-preferencing in theory, critics can now point at a specific set of answers and ask why no airline, no travel agent, and no comparison site received a single link.
Flights are not a random test case. Travel is one of the most commercial corners of search, and flight booking is a high-value transaction where the referrer earns real money. Airlines, online travel agents, and metasearch sites all depend on search referrals, and a shift in where flight links point can move serious revenue.
Google Flights is also a mature product with deep inventory and pricing data, which gives Google a ready answer to the criticism. The company can argue that it links to Google Flights because that is where the best flight data lives, not because it owns the property. That argument is plausible on its face, and it is also the argument Google has made in similar disputes before.
The practical question is whether users are getting the best answer or the most convenient answer for Google. A flight answer that links only to Google Flights may still be useful, and many users may prefer it. But a user who never sees an airline's own site, a smaller travel agent, or a price comparison service is making a choice with less information than a traditional search results page would have offered.
Self-preferencing is the idea that a platform which controls both the marketplace and a product sold in it will favour its own product. Regulators in Europe and the United States have spent years investigating whether Google does this in search, shopping, and travel. Google has consistently denied that it ranks its own services above better alternatives.
AI Mode changes the shape of that debate. In classic search, self-preferencing shows up as a ranking position or a prominent box. In an AI answer, it shows up as a link, or as the absence of links to everyone else. There is no page two to scroll to and no organic list to compare against. The answer is the whole result, which makes the choice of links inside it more consequential than a single ranking slot ever was.
That suggests the scrutiny of AI Mode linking will only grow. Every vertical where Google owns a destination product, from flights to hotels to jobs to shopping, is a candidate for the same test Schwartz ran on flights. If the pattern repeats across verticals, the debate moves from a single anecdote to a systemic question.
For SEOs in travel, the implication is direct. If AI Mode becomes a major way users research flights, and its answers link only to Google Flights, then ranking well in classic search may no longer be enough to earn the click. Travel sites need to track their visibility inside AI Mode answers separately from their classic rankings, and to measure what share of AI Mode responses cite or link to them at all.
For link builders, the lesson is broader. The value of a link depends on where it sits, and a link inside an AI answer may soon be worth more than a link on a results page nobody scrolls to. Earning those links means being the source the AI trusts for a specific fact, price, or comparison. That is a different skill from classic link building, and it rewards pages that are precise, current, and easy for an AI to parse.
The practical takeaway is to run your own version of this test in your niche. Ask AI Mode the questions your customers ask, note every link in the answers, and record who gets cited and who gets ignored. That record is the evidence base for your next content and outreach plan, and it is far more useful than guessing how AI Mode treats your vertical.
Sources: Search Engine Roundtable, November 11, 2026.
Related reading on this site: OpenAI Launches Businesses in ChatGPT With Free Profiles and AI Agents, Google Tests an AI Overview Local Pack Without Buttons, Reviews or Photos, Google's Search Volume Is About 5x ChatGPT's, OpenAI Says, AI Overviews Appear on 70% of Everyday UK Searches, Sayu Study Finds, Google Search Ranking Volatility Spikes Again Around November 8-9, and Google AI Overviews Cost French Sites Up to 69% of Clicks — But Not Every Site.
Linkbaits helps you create content that earns citations and backlinks. Start with the toolkit and build your first linkable asset this week.
Start free trial