Japan Airlines is getting fewer visits to its website, and different visitors when they arrive. Daily visits fell 10% in 2026 through early July against the same period in 2025, according to Minako Kent, the airline’s VP of global brand marketing. Those who do come through are more often loyal customers arriving later in their research, already informed. Google’s AI Overviews answer questions without a click, and ChatGPT, Reddit and YouTube shape decisions earlier.
JAL’s response is a metric called share of model — how often a large language model references a brand’s products and services, and the sentiment around the brand compared with rivals. Share of Model is also the name of a tool from JAL’s agency, Jellyfish, which reads how AI describes a brand and feeds that into paid media buying. On Performance Max, Jellyfish used it to add more than 150 targeted search themes across six ad asset groups and shift spend toward JAL’s highest-value routes, including JFK and San Diego. A Jellyfish case study puts the result at a 56% lift in conversion rate, 14% better return on ad spend, a 21% drop in cost per acquisition, and 12.5% less media spend.
The same data showed JAL performing unevenly across models. “Gemini has the highest brand score compared to all other LLMs, so ChatGPT is actually one of the lowest,” Kent said. The airline plans its first paid advertising on ChatGPT in August. OpenAI launched ads in ChatGPT in 2026, and reporting for advertisers remains limited, leaving brands to third-party tools. Jellyfish told clients it would extend Share of Model analysis beyond Google and Performance Max to ChatGPT, Reddit, TikTok and YouTube.
Reddit is a separate problem. JAL isn’t described badly there; it barely appears — and Reddit is one of the top sources models draw on for travel questions, particularly premium seating, which JAL is pushing harder.
John Dawson, Jellyfish’s VP of strategy, said the two work together: “The biggest action that happens on either side of using ChatGPT is Google Search. ChatGPT ads don’t work in isolation.” How much paid activity on one platform moves a brand’s standing on another is still unclear — Jellyfish has early signals in some categories, nothing it will call a pattern. Kent’s constraint is simpler. “We have a certain amount of budget, right? And we are not increasing because of LLM.”
Read more at Ad Age.