Comscore says ChatGPT is losing ground: what AI share of voice actually means for CMOs

The concept at stake here is AI share of voiceshare of voiceYour brand's share of total advertising or conversation volume in your category, measured against competitors over a defined period.View full definition →: which brands get cited by AI systems when users ask questions, and how that citation landscape shifts as multiple competing AI platforms mature. Most marketing teams still treat this as an SEOSEOSearch Engine Optimization: the practice of improving your pages' natural (unpaid) rankings in search engine results pages to attract more organic traffic.View full definition → footnote. It is not. It is quickly becoming a primary distribution channel for brand discovery, and the data now shows it is not a single channel with one dominant player.

Comscore data reported by Digiday in 2026 shows ChatGPT losing meaningful ground to Gemini and Claude as the three platforms diverge in how they select, cite, and present brand and publisher content. That splintering is the core problem for CMOs: strategies built around a single AI gatekeeping discovery were always fragile, and the fragility is now visible in the numbers.

Why AI citations differ from search sharesearch shareThe proportion of searches in your category that mention your brand, a leading indicator of market share.View full definition → of voice

Traditional share of voice lived in search rankings and media impressionimpressionThe total number of times an ad or piece of content is displayed, regardless of clicks. Each display counts as one impression, even to the same person.View full definition → data. A brand could pull a monthly Semrush report, check its visibility scores against competitors, and mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → a content calendar accordingly. AI citations operate differently: they are generated responses, not ranked lists. A user asking Claude "what CRMCRMCustomer Relationship Management: software and strategy to manage and analyse customer interactions throughout their lifecycle.View full definition → should a mid-market SaaS company use?" does not get a page of ten blue links. They get a synthesised answer, and either your brand is in that answer or it is not.

The stakes compound because AI-referred traffic behaves differently to organic search trafficorganic search trafficVisitors arriving via non-paid (unpaid) search engine results, earned through content relevance and SEO rather than advertising spend.View full definition →. Users arriving from an AI citation are further along a decision process. They have already asked a qualifying question. That makes citation-driven visitors disproportionately valuable even when raw traffic volumes remain modest compared to traditional search.

The Comscore finding adds a layer of operational complexity that most marketing functions have not planned for. Optimising for ChatGPT's citation logic is a different task to optimising for Gemini's, which is trained differently, retrieves content differently, and operates inside Google's own product surface where brand authority signals from the broader Google index carry more weight. Claude, built by Anthropic, applies different source evaluation criteria again. A CMO who directed the SEO team to "get us cited by AI" in early 2025 probably imagined a single target. By mid-2026, there are at least three with meaningfully different mechanics, and the gap between them is widening.

How do ChatGPT, Gemini and Claude pick which brands to cite?

Each major AI system generates responses by drawing on a combination of training data, real-time retrieval (where enabled), and internal ranking logic that weighs source credibility. The divergence between ChatGPT, Gemini, and Claude comes from several places.

Gemini benefits from Google's established authority index. A brand that ranks well in organic search and carries strong E-E-A-T signals (experience, expertise, authoritativeness, trustworthiness) has a structural advantage in Gemini citations because the system can draw on the same signals Google Search already uses. This is why publishers who have invested in bylined expert content, structured data, and consistent topical authority tend to perform better in Gemini than in the other two.

ChatGPT, particularly in its default non-retrieval mode, leans more heavily on training data patterns. Brands that appeared frequently and positively in high-quality web content before the training cutoff have an embedded advantage that is slow to shift. New brands or repositioned brands face a lag problem: even good recent content may not move the needle quickly because the underlying model's knowledge is frozen until a new training cycle.

Claude, developed by Anthropic, has shown a tendency to weight primary sources and direct citations more heavily. HubSpot (a CRM and marketing automationmarketing automationUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → vendor, so treat their findings with the appropriate scepticism) ran experiments in 2026 suggesting that structured LinkedIn content from credible individual voices improved Claude's citation rates, though the mechanism is not fully transparent and the data comes from a vendor with a clear interest in promoting content activity on professional platforms.

A concrete example: a B2B logistics software company that invests in one detailed, expert-authored white paper on supply chain risk modelling, indexed properly with structured markup, is more likely to surface in Gemini citations than a competitor who publishes fifteen short-form blog posts on the same topic. In ChatGPT, brand recognitionbrand recognitionThe degree to which your target audience recognises or recalls your brand, either prompted or unprompted. It measures how present your brand is in people's minds.View full definition → built over years of consistent publishing may outweigh either. In Claude, clear attributionattributionA framework for assigning credit to the touchpoints that contributed to a conversion, so you can measure which channels and interactions actually drive results.View full definition → of expertise to named individuals within the content can matter more than either of the above.

Understandinghow AI systems select and surface brand contentbrand contentA strategy of creating and distributing valuable content to attract, engage and retain a defined target audience, rather than pitching products directly.View full definition → requires thinking about each platform as a distinct audience with different evaluation criteria, not a single monolithic "AI search" channel.

Should CMOs shift budget to AI citation optimisation now?

The Comscore data justifies attention. It does not justify an immediate full-scale pivot of budget and resources toward AI citation optimisation, for several reasons.

Attribution remains genuinely difficult. When a user reads a Claude response that mentions your brand and then visits your site three days later via direct traffic or a branded search, most analytics stacks will not connect those events. Server-side tracking improvements help with what happens after the click, but the citation itself leaves no standard UTM trail. You are optimising for an outcome you cannot yet measure cleanly in most organisations.

The platforms are not stable. Google's antitrust proceedings (regulators opened formal proceedings targeting Google's programmatic ad infrastructure in 2025 and 2026, as covered by Digiday) create structural uncertainty about Gemini's distribution and commercial model. ChatGPT's retrieval capabilities and citation logic have changed materially across several product updates. Building heavy infrastructure around any one platform's current citation behaviour is a bet on something that could shift within a product cycle.

What does hold across all three platforms is content quality and structural clarity. Expert-authored content, properly marked up, attributed to credible voices, organised in topical depth rather than surface breadth, performs better in AI citations across all three systems than thin, high-volume content production does. That is also whata rigorous approach to content strategycontent strategyA strategy of creating and distributing valuable content to attract, engage and retain a defined target audience, rather than pitching products directly.View full definition → has always recommended, so the investment is not wasted if the AI citation model changes again.

The CMO's near-term job is to start measuring AI visibility systematically, split by platform, and build the internal case for treating citations as a tracked metric rather than an ambient outcome. Semrush (which sells SEO and AI visibility tooling, so assess their methodology accordingly) released an AI visibility audit framework in 2026 that offers one operational starting point, though independent validation of the methodology is worth seeking before it becomes a board-level KPIKPIKey Performance Indicator, a measurable value that shows how effectively you're achieving a specific objective, tracked over time against a target.View full definition →.

The fragmentation Comscore has documented is not a temporary phase before one platform wins. Multiple AI discovery channels will coexist, with different citation logic and different user bases. A brand that builds authority signals broad enough to perform across all three is better positioned than one that treats this as an optimisation task for a single platform. Start there.