The clearest sign that AI attribution is broken in 2026: 91% of AI-generated citations appear in only one platform, never simultaneously across ChatGPT, Perplexity, and Google’s AI Overviews, according to Kevin Indig’s AI Halftime Report (H1 2026), as reported by Greg Jarboe at Search Engine Journal. If your team tracks only one AI engine, you are, by that data, capturing only a fraction of your brand’s real AI footprint.
Indig’s report opens with a thesis that held across every storyline he tracked through June 2026: AI’s impact kept growing faster than anyone’s ability to measure it. That gap between impact and measurement is, in his framing, the central story of the first half of the year, more significant than any single product launch or earnings call.
Why does the AI attribution gap matter for brands right now?
Because trust, not position, appears to be the dominant factor when consumers pick from an AI-generated shortlist. Indig’s research found roughly three out of four consumers choose the top result in an AI answer, unless a brand they already recognize appears elsewhere on that list, in which case they select the familiar name. Brand awareness now functions as a de facto ranking override in AI-mediated results.
Software stock performance in H1 2026 reflected the same dynamic. According to Indig’s report, valuations fell by close to 30% over the period, tracking not with how companies actually performed financially but with how the market perceived each company’s exposure to AI disruption. The bottom quartile of software stocks dragged the sector down, while the median and top quartile outperformed the broader ETF. The selloff tracked narrative, not fundamentals.
What does the citation data actually show?
The 91% fragmentation figure is the most strategically significant number in Indig’s H1 2026 report. Nine out of ten citations appear in only one of the three major AI answer surfaces, never in two or more. The practical implication: prompt tracking (monitoring what an AI says in response to a specific question) needs to function more like polling research than like the rank tracking that SEOs have relied on for the past two decades.
- 91% of AI citations appear on only one platform (ChatGPT, Perplexity, or AI Overviews), per Kevin Indig’s H1 2026 report
- Google Search Console data is reportedly 75% incomplete for the new AI-search landscape, per the same report
- Brand mentions, not citation counts, correlate more closely with real business outcomes in Indig’s analysis
- ChatGPT’s share of the AI assistant market slid from 78% to 56% between July 2025 and July 2026, per Indig’s data
- Gemini climbed from 15% to 30% over the same period; Claude grew from 2% to 10%
The platform-share shift compounds the measurement problem. A brand tracking only ChatGPT in July 2025 captured roughly 78% of AI assistant usage. The same single-platform approach in July 2026 captures only 56% of that market and misses a Gemini that doubled its share in twelve months.
Did AI actually drive H1 2026 layoffs?
Challenger, Gray & Christmas data, cited in Indig’s report, found AI attributed as the reason behind more than 87,000 job cuts through May 2026, roughly one-fifth of all layoffs over that period. Indig’s own analysis, consistent with predictions he made in his H1 2025 report, argues AI was mostly a convenient label for cuts primarily driven by pandemic over-hiring and capital expenditure discipline. Several companies that publicly attributed layoffs to AI reportedly rehired quietly in the same months.
Meta’s internal spending data, also cited in the report, adds context. Meta engineers reportedly consumed 73.7 trillion tokens (the unit AI models use to process text, roughly equivalent to individual word-chunks) in a single month, running against an internal leaderboard that ranked more than 85,000 employees by consumption volume. The company shut the leaderboard down in April, after CFOs identified that annual token budgets had been exhausted in roughly four months with no clear return attributed to the spending.
What are publishers doing about AI content use?
Publishers moved the dispute from analytics tools to courts and regulators in H1 2026. According to Indig’s report:
- A Munich court ruled Google liable for false statements generated by AI Overviews
- 400 newspapers sued OpenAI and Microsoft over unauthorized use of their content
- The UK’s Competition and Markets Authority (CMA, the UK’s independent competition regulator) ordered Google to give publishers more control and transparency over how their content is used in AI search, including opt-out rights from AI Overviews, AI Mode, and Discover summaries
Each dispute centers on the same question: who controls content access and attribution once a human is no longer the one clicking through to a source.
What this means for AI-search visibility
Here is the plainest way to read what H1 2026 revealed: if a potential customer asks an AI assistant about your product category and 91% of citations appear on only one platform, then a brand appearing in ChatGPT almost certainly is not appearing in Gemini or AI Overviews at the same time. That is not a performance failure. It is a structural measurement gap, and most dashboards are not built to show it.
In Hingewise’s reading of Indig’s report and Jarboe’s analysis, three implications stand out.
First, citation counts are a partial metric. If brands that earn citations in one engine rarely earn them in another, a high citation count in one platform tells you little about cross-platform brand presence. Mention frequency and recommendation rank across multiple engines, treating results like a polling sample, is closer to what the data actually supports as a meaningful measure.
Second, the platform-share shift is not a background trend. A measurement approach built around ChatGPT’s 78% dominance in mid-2025 faces a meaningfully different competitive landscape in mid-2026, with Gemini at 30% and Claude at 10%. Model choice has become a business risk variable, not a technical preference.
Third, and this is a Hingewise reading that goes beyond what the sources explicitly frame: the distinction Indig previews for H2 2026, between AI that answers questions (which he calls intelligence) and AI that takes actions on a user’s behalf (which he calls agency), represents the next visible gap in AI attribution. Being cited in an AI answer still requires a human to decide and click. Being recognized by an AI agent (software that can act autonomously on behalf of a user) as a trusted, transactable option means the AI itself may complete a booking, a purchase, or a referral without any human click at all. Brands whose product data, pricing, and checkout flows are not structured for agent access risk becoming invisible at the exact moment AI is most commercially active. That is a different kind of attribution gap entirely, and no tool currently tracking AI citation or mention data is built to surface it.
The H2 signal worth watching is whether the regulatory pressure publishers applied to content attribution produces any spillover toward standards for brand attribution in AI search, or whether the gap between AI’s commercial impact and our ability to measure it widens further before any standardization emerges.
Before you next report on AI search performance: a quick pre-check
- Are you tracking brand mentions across at least three AI platforms, not just citations from one?
- Is your prompt panel large enough to function like a polling sample rather than a handful of spot checks?
- Are you accounting for the 75% incompleteness Indig identifies in Google Search Console’s AI-search data?
- Does your reporting distinguish between citation counts and mention frequency plus recommendation rank?
- Have you updated your platform-mix assumptions since mid-2025, given Gemini’s share growth and ChatGPT’s decline?
- Is your product data and pricing structured so an AI agent could, in principle, access and act on it without a human click?
- Are your opt-in or opt-out positions for AI Overviews and AI Mode reviewed in light of the UK CMA’s new publisher transparency requirements?
Sources: Greg Jarboe, “AI’s Impact Is Outrunning Measurement: The Trust And Attribution Gap Facing Brands,” Search Engine Journal, 2026 (searchenginejournal.com), referencing Kevin Indig’s AI Halftime Report, H1 2026; Challenger, Gray & Christmas, cited in same.
Author: Lam Nguyen · Hingewise