In a discussion hosted by Ezra Eeman, as part of the WAN-IFRA AI in Media Initiative he leads – and the first in a new series called “Publishers and Platforms” – Varun Shetty, VP, Media Partnerships at OpenAI, described their 2026 priorities across personalisation, proactive experiences and the evolution of ChatGPT Search, an apps ecosystem designed for account linking and action-taking, and experiments with monetisation.

The headline ambition for 2026: personalisation + proactivity

“If you had a human assistant, and you could only get information from them by asking them questions, I think you would be unsatisfied with that,” says Shetty. “What you would want is your assistant to deeply understand you – your preferences, your goals, your habits – and be able to come to you with information that helps you accomplish all of those things.”

From that observation, Shetty draws a direct line to 2026: “As we look to 2026, a very big opportunity for ChatGPT is to increase personalisation and make it easier for the people who use ChatGPT to take actions and accomplish their goals.”

That’s not abstract positioning – there are already “seeds” of this approach in recently launched agentic features and more verticalised experiences.

Pulse: A ‘daily digest’ that could matter for news

Pulse is a good example of a proactive experience. It’s an agentic product that searches overnight for information related to a user’s interests and brings it back.

Shetty described it as “essentially a daily digest that is served to you within ChatGPT” and that “pushes you information” based on what you’ve been chatting about recently.

But Pulse also highlights a constraint shaping OpenAI’s roadmap: cost. Shetty called it “a relatively compute-intensive product… requiring a lot of GPUs,” which is why it is currently limited “to our pro users in the US,” while OpenAI invests in compute power and inference in other regions to expand access.

Search: grounding, citations, link-outs – and a more ‘bespoke’ news experience

Shetty described ChatGPT Search as OpenAI’s answer to a basic limitation of early ChatGPT: it couldn’t reliably handle “recent events” queries.

During 2024 and 2025, OpenAI launched a search experience that uses RAG to return answers “grounded in information from high-quality trusted publishers.”

From a publisher perspective, Shetty stressed what he called “core non-negotiables”:
• proper citation
• as many link-out experiences as possible
• and an evolution towards a search experience that is “more customised, personalised, more visual”.

He acknowledged a structural tension: ChatGPT is not a classic “10 blue links” interface.

“People aren’t just seeing links, they’re seeing answers. So the likelihood that they click through is naturally going to be lower than if you just see a link,” he said – while adding there is “headroom” to improve click-through.

OpenAI also intends to differentiate the “news experience” within ChatGPT from other verticals: “We can do more interesting things to have a bespoke news experience within ChatGPT that looks different than our commerce experience, that looks different than our local experience,” he said.

WAN-IFRA Members can replay the webinar on our Knowledge Hub by clicking here.

Shetty emphasised that publishers have explicit control over inclusion in ChatGPT Search.

“If you do not want to be surfaced in ChatGPT …  you can very simply do that … by editing your robots.txt file and preventing our crawlers from going to your site,” he said, referring to blocking OpenAI’s bots.

He framed that as a deliberate design choice – “important to allow for that opt-out”– and added that it creates “pressure on OpenAI to figure out how to create enough value for you to be drawn in.”

Shetty also suggested publishers could experiment temporarily by allowing the search bot, which “only allows us to display your content” and “does not allow us to train on your content,” in order to evaluate incremental outcomes.

During the webinar, participants asked for more transparency: impressions, categories of queries, what drives CTR, and how traffic behaves.

Shetty confirmed that OpenAI is thinking about it: “My hope is that over the course of this year, we can build the right metrics and dashboard experience so that people can understand more deeply exactly how the ecosystem is working.”

This matters because OpenAI’s pitch to publishers isn’t only “trust us” – it’s increasingly “measure it.”

Apps: An ‘open platform’ and the subscription opportunity

Shetty repeatedly returned to the idea that Search is only one part of the value proposition.

The “remaining parts,” he said, “centre on things like connectors, apps, and Pulse,” which he framed as “mid and lower funnel” ways for publishers to engage readers when they are away from their own platforms.

OpenAI’s apps ecosystem is meant to enable specialised actions that ChatGPT alone won’t do. OpenAI “announced this in November, we had some early launches in December, and now it’s an open platform,” Shetty explained.

He used Spotify as an analogy: a user can connect accounts, ask for a tailored playlist, and ChatGPT and Spotify together complete the action. For news, he said, OpenAI wants to explore a “third-party app experience”: if ChatGPT knows a user is a subscriber because their account is linked, that can unlock “personalisation and other opportunities.”

Mapping ChatGPT to the publisher funnel: traffic → engagement → conversion

When asked directly about business models for publishers – beyond traffic – Shetty described OpenAI’s current approach as mapping to existing publisher economics: “Can we drive traffic, and then can we drive deeper engagement, and then potentially subscription and conversion?”

He pointed to OpenAI’s commerce product as a foundation – currently focused on physical goods – but suggested it “could be extensible to digital goods as well.”

Then he made the publisher tie-in explicit: “So if we know that you are reading X publication, can we upsell you to a subscription? That’s not out of the realm of possibility for us … and can we do that pre-click … [or] post-click also?”

At the same time, Shetty said OpenAI is not currently considering certain industry proposals: “content marketplaces … paying on crawling … that is not something that we are exploring right now.”

When asked about initiatives like Cloudflare’s “pay-per crawl” and similar efforts, he reiterated that OpenAI has relied on robots.txt as the control mechanism, and said what OpenAI needs is “more advanced and sophisticated” value to entice opt-in—where “a lot of our investments are happening this year.”

Monetisation for OpenAI: subscriptions + enterprise + API… and advertising experiments (with ‘principles’)

Shetty described OpenAI’s monetisation mix as:

• Enterprise products (ChatGPT Enterprise / Business)
• The API for developers
• Consumer subscriptions – while stating, “we are going to experiment with advertising.”

“We do think it’s important that we figure out how to monetise our free user base… We want technology … in as many hands as possible … advertising can be a very powerful business model to make that happen,” he added.

On 16 January, the day after the webinar, OpenAI made an official announcement about its advertising intentions. In the coming weeks, the company will test advertising in the US on its free product and on a low-cost subscription tier, ChatGPT Go, which is launching in the US for $8 per month with expanded features (messaging, image generation, file uploads and memory) and is already available in 171 countries. This will be for logged-in adults on the Free and Go tiers. The Pro, Business and Enterprise solutions will remain ad-free.

OpenAI says it will follow clear principles: ads will not influence answers, will be clearly separated and labelled, and user conversations and data will not be sold to advertisers. Users will have control, including the ability to turn off personalisation and clear ad-related data.

Ads will initially appear at the bottom of answers when relevant to the conversation, with options to see why an ad is shown or to dismiss it.

During testing, ads will not appear near sensitive or regulated topics such as health, mental health or politics.

Business adoption: ‘Unlock economic value’ and the OpenAI Academy for news

Shetty described OpenAI’s second major priority as “unlocking economic value for business,” referencing improvements in agents and deployment inside organisations to “amplify their impact”.

For news organisations specifically, he split the work into two buckets:

• Consumer product work (how information flows through ChatGPT), and
• Business adoption (how publishers implement ChatGPT Enterprise and the API).
  

He referenced the “OpenAI Academy for news,” launched “at the end of the year,” intended to highlight use cases – an “index and a recipe book” – so that a news organisation can find examples of GPTs or integrations (fundraising, CMS, etc.) and learn from peers.

Local, independent, and diversity: personalisation as the mechanism

On local news visibility, Shetty said OpenAI is still learning what “local” means in ChatGPT Search behaviour (some queries are navigational, like finding a bakery), and he wants “more work to understand local news and the overall prevalence.”

On the risk that only big news brands get surfaced, Shetty tied diversity to the assistant mission: If the product serves the same publishers to everyone, “that’s probably not a very personalised super assistant,” so personalisation should “push us towards showing a diverse set of publishers… not just large ones, but also the more niche or specific ones.”

What’s coming to the interface: More visual, more multimedia

Shetty acknowledged a UX limitation: “especially search, it can feel like a wall of text coming at you,” and OpenAI will want to make it “more dynamic” with “more visuals” and move towards “a more multimedia experience” in search.

He didn’t commit to an implementation for video specifically – “I don’t yet know the implementation for how video will be involved” – but suggested this direction could “open up additional opportunities … for news publishers.”

‘Not a newsroom, but a pathway to trusted sources’

A final question raised a broader concern: the media’s role in public discourse.

Shetty responded by distinguishing OpenAI from a media company: “We are not a media company. We are not a newsroom.”

Instead, he framed OpenAI’s role as helping people get answers while connecting them with “sources and journalism that’s high quality and trusted,” enabling users to “deepen their understanding on specific topics.”