Inside Tripadvisor’s AI strategy

With over 1 billion reviews and millions of monthly users, Tripadvisor hosts one of the world’s largest travel and hospitality datasets – a scale at which consumer personalisation becomes “big business” and the cost of underutilising data becomes high.

Senior Director Sanchit Janeja knows this all too well - and while AI is helping with the task, the technology is not a panacea. He says each AI personalisation use-case must be carefully considered, with no single tool guaranteeing ROI.

“The question we are always asking is, how can we deploy technology, whilst making it profitable? So, it's not purely an engineering exercise.

“In any given context we are deciding whether we use traditional machine learning or generative AI and ensuring that whatever we use has a business purpose.

“Every decision I take, every ounce of engineering bandwidth that is spent is supposed to be North Star towards ROI.”

Ripple effects

Adding to that, the level at which the technology is deployed across all five tiers of the personalisation pyramid – user intent, recommendations, supply, search and content – affects the company’s broader technology ecosystem.

Sanchit said user intent, at the top of the pyramid, has major implications for frontend interfaces.

“The concept of a person using a website, mobile app, or Google search is evolving. So, a lot of companies are investing in new ways of interacting with users,” he said.

“For example, you or I might interact with Air BnB via website or mobile. But what happens if an agent is booking on your behalf, or you’re using ChatGPT to make your booking? You might tell ChatGPT you want four nights in a warm place, and ChatGPT in the back end would work with TripAdvisor. Based on your history, it would say, ‘I know that you like the Pacific Southwest, so maybe you want to go to Hawaii. I know you haven't been there, but you were searching for it a while back. I think you would like these options. Can I just book everything for you?’”

Pre-emptive transformation

In recognition that the entire user interface may change, some companies are designing their own interface, with the likes of Air BnB even creating an AI lab.

Dynamic front ends, such as Google Disco, are another popular workaround, in the wake of agentic commerce. This is where a new front end pops up, based on the consumer’s search.

Meanwhile, other companies are discussing the possibility that the front end may soon become obsolete.

“If you are working with an agent and if, let's say, Tripadvisor has its own agent or an MCP ecosystem exposed, then the agent is essentially talking to another agent and you don't even need a front end.

“In this case, we might need to personalise a front end from a form factor - maybe a mobile or web, or something like that.”

Content personalisation

When it comes to channeling technology spend, content is another popular area of focus on the personalisation pyramid.

Everything a consumer sees on a website, including after sales service, can be personalised - and Sanchit believes doing so is key to maintaining the value proposition of any ecommerce platform.

“Large language models are doing a lot of development – and they are getting cheap. The truly unique differentiating factor is your supply, your operations, as well as the website personalisation that you can offer.”

A moving target

With the AI scene constantly evolving, Sanchit said building a technology strategy is a moving target, acknowledging the high degree of risk in deciding and working on a vision that could become obsolete within years.

Regulation, too, is evolving at pace, with countries like China putting borders around its domestically farmed data, via its Personal Information Protection Law (PIPL), which prevents raw data from being transferred overseas.

Virtually every country also has its own measures to prevent bias and data drift.

“We are heavily bound by regulation and that is in a state of flux, so of course this presents a challenge as we figure out the best way to optimize our data for personalisation tasks,” Sanchit said.

Will AI threaten future ecommerce?

As technology gets more advanced, there are concerns it could cross the line from aide to competitor, outperforming the very companies they are deployed to serve. However, Sanchit does not believe Tripadvisor’s value proposition is under threat.

“A lot of large language models – OpenAI for example - have publicly stated they cannot replace e-commerce operations, such as booking.com or Tripadvisor, because they cannot do payments, aftersales support etcetera. So, this somewhat protects us from getting wiped out. Compare it to software as a service, which is facing a lot of headwinds due to AI.

“In e-commerce, our strongest value proposition is our operations. So we have a role to play in this LLM world. We just have to modernise ourselves accordingly. And this is where we are primarily focusing our investment on.”

Further insight

Sharing more about his work in this space, Sanchit will join an expert line-up of speakers at the upcoming AI Summit.

This year’s event will be held 7-9 September 2026 at the Melbourne Convention & Exhibition Centre.

Learn more and register your tickets here.