Exclusive: Comscore brings generative AI measurement to India

New Delhi: Comscore has brought its generative AI measurement capabilities to India, promising publishers and brands a clearer view of how audiences use AI assistants, when AI answers appear in search, and which domains those answers cite. 

In an exclusive conversation with BestMediaInfo during her India visit, Smriti Sharma, Senior Vice-President, Custom Solutions at Comscore, said the company is “bringing a capability” built this year into the Indian market so stakeholders can track discovery in the AI era.

Sharma framed the timing around a simple shift in online behaviour. People often receive a complete answer before they see a list of links, which depresses click-throughs and changes how attention moves. “Previously, you used to search and click a list of sites. Now the query is often satisfied by an automated answer. That journey is changing, and our job is to help publishers and marketers understand it and adapt,” she said.

What Comscore is launching in India

Gen AI Visitation. This measures traffic to AI tools, covering 100-plus entities on desktop and 50-plus on mobile, grouped into 19 categories including assistants, text, images, audio, video, design and coding. “We’ve created categories so an image tool is compared with image tools, a video generator with video generators. It makes the comparison fair,” Sharma said.

AI Summary Detection in Search. A new flag in Comscore’s search data indicates whether a Google AI Overview or a Bing Copilot Answer was present when a query was made or a click occurred. It enables analysis of the share of queries that trigger AI summaries, the categories most affected, and how the presence of summaries correlates with click-through behaviour. The signal is available for desktop on Google and Bing, with international feasibility on request.

Gen AI Prompts & Responses. This captures anonymised prompt–response pairs from leading assistants to reveal actual intents and multi-turn behaviour. Coverage today includes Copilot, Gemini and Perplexity, with ChatGPT in development. Historical data is available for Copilot and Gemini and for Perplexity from April 2025. “From the start of the session to the end, it is your customer journey with AI. We track the questions, the responses and whether AI was effective,” Sharma said.

Sharma added that, beyond visitation and prompts, Comscore is also building a citation view. “When an automated summary says ‘cited by,’ it drives intent. We are evolving a citation report alongside visitation,” she said.

Why this matters for India

Comscore argued that AI usage is now a core metric. As AI shapes search, recommendations and creative tools, tracking adoption helps marketers see which platforms drive engagement and where opportunities exist. “In a fragmented environment, the need for trusted, cross-platform measurement is greater than ever,” Sharma said, noting India’s fast growth in digital and connected TV.

Early signals point to rapid uptake. In India, Comscore observed new AI-assistant visitors across platforms over the last year, with mobile visitation up 153% and desktop up 41% between November 2024 and September 2025. The first AI touchpoint is increasingly mobile.

Relevance for marketers

“Marketers need to know where their audience actually is. If you are targeting 40-plus, our data may show they spend more time with ChatGPT than Copilot. That changes where you place ads and what creative you make. The suite gives a single view of how different segments use AI tools across devices and how adoption is evolving. AI is no longer a separate technology. It is part of daily life, so planning and measurement must reflect that,” Sharma said.

She also pushed back on the idea that AI shortens consideration. “Customer journeys with AI are expanding, not shrinking. For marketers, that means more opportunities to reach people, as long as they know the touchpoints.”

From a brand lens, Sharma said the citation and visitation report enables true cause-and-effect analysis. “You can measure PR and earned-media impact by seeing which articles and which news sites cite your brand, and how many site or app visits are associated with those citations. You can trend citations and visits over days to plan PR and decide where you want traffic from. It gives you evidence for leadership that a press release or campaign actually moved an audience. You can diagnose gaps, high citation with low visitation or the reverse, and course-correct. And yes, you can benchmark against competitors to see how you stack up.”

How publishers can use it

For publishers, the practical questions are straightforward: Are we being cited in AI answers, how often, and on which topics? When AI summaries appear, does traffic still come through? Which assistants are our audiences using?

Sharma’s advice is blunt. “If you do not change, you will not be cited anymore,” she said, highlighting AEO, AI engine optimisation, as a necessary complement to SEO. 

In her framing, AEO is about making content answer-friendly, ensuring page code and structure are modular enough for large models to parse, and monitoring mentions inside AI overviews.

The visitation dataset lets publishers see growth or declines across categories and tools so editorial and product teams can align coverage to actual behaviour. The search flag helps quantify the impact of AI summaries on click-through, including the share of searches that end without a further click. Prompts and responses can reveal where content is missing. If assistants frequently answer with sources outside a publisher’s domain on a beat the newsroom wants to own, that gap is visible. 

Sharma said these outputs will begin moving into syndicated rankers inside Comscore’s broader products so publishers can benchmark routinely rather than treat AI as an ad-hoc study. “We will start moving into syndicated, because that becomes the new rankers,” she said.

LIVE data for reach, frequency and attribution

Sharma also pointed to LIVE analytics for moments that cannot wait for the month-end. Where platforms share data, Comscore can provide deduplicated, cross-device reach and frequency for high-stakes events within roughly half a day. 

“Those are things we have done in the US, and we are open to doing them in India,” she said. 

Sharma continued, “Indian audiences watch live across TV, apps, YouTube and social. LIVE Comscore shows deduplicated cross-platform reach, not just impressions. That tells advertisers whether they are reaching new people or the same ones and helps tune reach and frequency across devices. Our LIVE analytics read ad exposure and response in real time and feed AI-driven attribution, so teams optimise during the event, not after. This will be central to CTV and OTT, elections, sports and festival campaigns, and retail media. It turns Comscore from measurement to a real-time decision intelligence partner.”

Methodology and privacy

Sharma was explicit about how Comscore collects AI signals. The work is anchored in opt-in desktop and mobile panels, merged with tagged datasets. The company does not rely on synthetic prompts or scraping to fabricate behaviour. “It is our opt-in-based panel. People have opted in. Panel provides quality and tags provide quantity,” she said.

On privacy, the promise is that no personally identifiable information is shared. “I will know the age group, region and occupation, but I will not know it is you,” Sharma said. She added that this structure aligns with India’s data-protection regime because participation is voluntary and the analysis is aggregated.

Scope, caveats and what is next

The AI-in-Search flag currently covers desktop for Google and Bing, with feasibility checks for other markets. It can show when AI summaries appear and how that presence maps to clicks, including the share of queries that terminate at the answer.

The Prompts & Responses feed begins with Copilot, Gemini and Perplexity and will expand. Limitations include ChatGPT coverage being added, platform-specific start months for history, and extraction limits that may render portions of some prompts or responses partially legible. The feed is not intended to infer the overall assistant market share.

Sharma said the capability was built in September and recently sold to early clients. The company is adding more entities and features and “bringing awareness” that the suite exists. In India, the Custom team expects to layer on adjacent services that marketers already recognise, customer journeys, benchmarkers, brand-lift studies and brand monitors, delivered on the same underlying panel logic.

What sets Comscore apart

“What sets Comscore apart is the source and quality of our data. We measure real behaviour from our own opt-in desktop and mobile panels, combined with tagged datasets. Others often rely on synthetic prompts or synthetic panels that simulate usage. Ours are privacy-controlled, validated panels that give deduplicated, cross-device reach. That means publishers and marketers can finally see how they are being surfaced inside AI tools with confidence,” Sharma said. “Comscore is the only one with cross-platform reach. We remove duplicates. It is not about impressions; it is about actual reach after de-duplication.”

How it will be sold

Sharma said Comscore is delivering these AI datasets as monthly reports, with options to slice by month, year-over-year or full-year series. As the market matures, parts of the AI suite will flow into syndicated products so that tracking becomes routine. When a client needs custom cuts, such as isolating a sector, mapping a specific campaign, or combining datasets with first-party feeds, the Custom team engages. “Because it is not a license, you can pick and choose based on what fits you,” she said.

Comscore’s move gives India’s publishers and marketers a third-party way to see what AI is doing to discovery. It connects three questions that have, until now, been answered separately: where people go, when a search turns into an answer, and what people ask assistants for. The data will not end debates about AI fairness or traffic cannibalisation. It does, however, create a common yardstick to plan against in a market where AI adoption is rising quickly and mobile behaviour leads the way. As Sharma put it, the companies that adapt first will be the ones models cite most.