A test published over the weekend produced a stark result: every single AI Mode response to travel queries linked exclusively to Google Flights. Not a mix of sources. Not a blend of airline sites, OTAs, and aggregators. Just Google's own product, every time.
This is not a minor UX quirk. It is a signal about how AI Mode allocates citation links - and what that means for brands who have invested in content, schema, and authority signals specifically to earn visibility in AI search.
What the Test Actually Showed
The pattern was consistent: 100% of outbound links in AI Mode travel responses pointed to Google Flights. If that holds across a broader sample, it suggests AI Mode is not selecting citations through the same editorial weighting that governs AI Overviews or the organic results beneath them.
Whether this is a deliberate product decision, an artefact of how AI Mode is trained on structured data sources, or something that shifts query by query is not yet clear. What is clear is that for travel queries at least, third-party brands are not getting the link.
The source article notes this will likely attract attention from Digital Markets Act regulators in Europe. That is a reasonable expectation. Self-preferencing in search results has been a central concern of EU competition enforcement, and an AI-generated surface that consistently routes users to a Google-owned product is precisely the kind of behaviour the DMA was designed to address.
The Difference Between Being Cited and Being Linked
There is an important distinction here that AEO practitioners need to hold onto. AI Mode can reference information from a third-party source without linking to it. A response might draw on content from an airline's fare pages or a travel publisher's guide while routing the click to Google Flights. The citation and the link are not the same thing.
For brands measuring AI search performance, this matters enormously. If your content is informing the response but not receiving the click, your GA4 data and CRM attribution will show nothing. You would have contributed to the answer and received zero measurable traffic or leads from it.
This is why tracking referral traffic from AI surfaces - using UTM parameters where possible, referrer analysis, and direct session monitoring - is not optional. Without it, you cannot know whether AI Mode is using your content without crediting you with the visit.
Verticals Where This Risk Is Structural
Travel is not an isolated case. Google has built verticals across hotels, flights, shopping, and local services. In each of these, there is a Google-owned surface that could theoretically absorb the link at the point of AI Mode response - Google Hotels, Google Shopping, Google Flights, Google Maps. These are categories where many brands have been building AEO programmes.
The strategic implication is not to abandon those programmes, but to diversify where you are building visibility. ChatGPT and Perplexity do not carry the same vertically integrated product conflicts. When a user asks Perplexity about flights, there is no Perplexity Flights competing for the link. That changes the citation dynamic considerably.
For brands in sectors where Google has a competing product, the case for building presence across multiple AI search surfaces - not just Google - becomes a risk management argument, not just a reach argument. If AI Mode routes links internally in those categories, your content needs to be citation-worthy on platforms without that structural conflict.
What This Changes About How You Measure AI Visibility
If self-preferencing is real and consistent in AI Mode, it creates a measurement gap that standard reporting will not catch. A brand could be appearing in AI Mode responses, contributing to the informational content of those answers, and receiving no attributed sessions at all. The standard metric - AI-referred traffic in GA4 - would show a flat line.
This means measurement programmes need to include prompt-based auditing alongside traffic analysis. Regularly querying AI surfaces with the searches your customers are making, and recording whether your brand appears in the response, whether it receives a link, and what that link points to, gives you data that session-level reporting cannot.
If you are tracking AI-referred leads through to CRM and you operate in a Google vertical category, it is worth running a parallel audit of AI Mode responses to understand what share of citations are going to Google's own properties. That gap between content contribution and actual traffic is a business risk, not just a measurement inconvenience.
What Brands Should Do With This Information
First, do the test. Run your most commercially important queries through AI Mode and record what links appear. If you are in travel, finance, shopping, or local services, check whether Google's own products are absorbing the links your content might otherwise earn. Do this regularly, because this behaviour can change with model updates.
Second, weight your AEO investment accordingly. Content and entity signals that earn citations in Perplexity, ChatGPT, and Gemini are built on the same foundations as Google AI Overviews - structured content, authoritative sources, strong entity relationships, digital PR, and clear schema signals. That work is not wasted. But the platform weighting in your measurement should reflect where you can actually capture a link.
Third, use the regulatory context as a timing signal. The DMA is already a live constraint on Google's behaviour in Europe. If self-preferencing in AI Mode attracts regulatory scrutiny - which the source article suggests is plausible - the citation allocation may shift. Brands that have built strong content and entity signals now will be positioned to capture those links if and when the behaviour is forced to change.
The Broader Point About AI Citation Trust
One thing AEO programmes have always had to contend with is the gap between being a good answer and being the chosen answer. AI surfaces make editorial decisions - which sources to cite, which links to surface, how to frame the response. Those decisions are not always based purely on content quality.
What the Google Flights finding adds to that picture is the possibility that platform economics - not just content merit - are shaping which brands get the click. That is a materially different problem from optimising your schema or improving your content structure. It is a distribution problem, and it requires a distribution response: building presence across multiple AI surfaces rather than concentrating entirely on the one with the most users.
AI search visibility is not a single-platform problem. The brands that treat it as one are building on foundations that a single algorithmic or regulatory shift can undermine.