AI’s Impact Is Outrunning Measurement: The Trust And Attribution Gap Facing Brands

Aug 08, 2026 07:00 PM - 1 hour ago 1

In the first half of 2026, search behavior shifted nether our feet, package valuations cratered connected sentiment alternatively than proof, and companies blamed layoffs connected AI agelong earlier anyone could show the receipts.

Kevin Indig posted a nexus connected LinkedIn past week to his AI Halftime Report, H1 2026, and it is worthy noting that his H1 2025 study predicted some Google’s continued AI Mode rollout and the thought that AI layoffs were mostly a PR communicative alternatively than an operational reality. Both predictions held up, and the study opens pinch a statement that doubles arsenic the thesis for the full first half of the year: AI’s effect kept increasing faster than anyone’s expertise to measurement it. That spread betwixt effect and measurement is the existent communicative of H1 2026, much than immoderate azygous merchandise motorboat aliases net call.

Trust Became A Ranking Factor Readers Feel In Their Own Results

Indig’s investigation recovered that astir 3 retired of 4 consumers prime the apical consequence successful an AI shortlist, unless a brand they already trust appears anyplace other connected that list, successful which lawsuit they prime the trusted sanction instead. Software stocks fell by adjacent to 30% complete the period, and the diminution tracked almost wholly pinch really the marketplace perceived a company’s vulnerability to AI disruption alternatively than pinch really that institution really performed. The bottommost quartile of package stocks dragged the full assemblage down while the median and apical quartile outperformed the broader ETF, which tells you the selloff was a communicative astir narrative, not fundamentals.

Meta engineers reportedly burned done 73.7 trillion tokens successful a azygous period chasing an soul leaderboard that classed much than 85,000 labor by token consumption. Nobody could constituent to the return connected that spending, and Meta unopen the leaderboard down successful April erstwhile CFOs realized yearly token budgets had been blown done successful 4 months.

AI Took The Blame For Cuts It Did Not Make

Layoffs told a akin communicative of communicative racing up of proof. Challenger, Gray & Christmas recovered AI cited arsenic the logic down much than 87,000 occupation cuts done May, astir a 5th of each 2026 layoffs to that point. Indig’s ain reporting, going backmost to his H1 2025 predictions, argues AI is mostly a convenient mentation for cuts that were really astir pandemic over-hiring and superior expenditure discipline. Some of the aforesaid companies that blamed AI for layoffs turned astir and softly rehired.

My return (and I person been saying versions of this since spring) is that Indig has named thing structural alternatively than seasonal. Every 1 of his H1 storylines, the SaaS selloff, the AI layoff narrative, the traffic collapse publishers are now fighting successful court, comes backmost to the aforesaid guidelines cause. We built our measurement scoreboards for classical hunt years ago, and nary of them were designed to show thing arsenic nonlinear and probabilistic arsenic AI hunt has turned retired to be.

That nonlinear communicative shows up starkly successful Indig’s citation data. He recovered that 91% of citations look successful only 1 of ChatGPT, Perplexity, aliases AI Overviews, ne'er successful much than one. Prompt tracking, he argues, needs to behave much for illustration polling and focus-group investigation than for illustration the rank search SEOs person leaned connected for 2 decades. Brand mentions, not citations, correlate much intimately pinch existent business outcomes, since astir buyers attraction whether a marque shows up favorably crossed a sheet of prompts alternatively than whether 1 circumstantial citation landed successful 1 circumstantial answer. Google’s ain Search Console data is reportedly 75% incomplete for this caller landscape, truthful moreover practitioners who deliberation they are measuring cautiously are moving from a partial picture.

There was an upside to each that token burning. Indig points retired that the activity of usage astir Claude’s Opus releases pushed acold much group into daily AI use than immoderate trading run could have, and that surge fed maturation astatine infrastructure companies further down the stack, the benignant of second-order effect that seldom shows up successful a quarterly net telephone but shapes an full ecosystem anyway.

The supplier marketplace itself reshuffled conscionable arsenic overmuch arsenic the metrics did. ChatGPT’s share slid from 78% to 56% betwixt July 2025 and July 2026, while Gemini climbed from 15% to 30% and Claude grew from 2% to 10%. Model prime has go a genuine business consequence alternatively than a preference, which is 1 much adaptable that resists easy measurement.

Publishers Took The Fight To Courts And Regulators

Publishers spent H1 2026 fighting the aforesaid conflict successful a different arena. A Munich court ruled Google liable for mendacious statements generated by AI Overviews. Four 100 newspapers sued OpenAI and Microsoft complete unauthorized contented use. The UK’s Competition and Markets Authority ordered Google to springiness publishers much power and transparency complete really their contented is utilized successful AI search, including opt-outs from AI Overviews, AI Mode, and Discover summaries. Every 1 of those fights is really an statement complete who gets to power entree and attribution erstwhile a quality is nary longer the 1 clicking through.

The Second Half Splits Intelligence From Agency

Indig closes his study pinch a preview of the 2nd half that deserves arsenic overmuch attraction arsenic thing successful the H1 recap. He argues H2 2026 will “separate intelligence from agency.” Model capacity keeps getting cheaper and much commoditized done open-weight competition, but the support to really enactment connected someone’s behalf, walk money, scope data, aliases impersonate a personification done an agent, is getting locked down harder by platforms, governments, costs networks, and users themselves. The patient lawsuits and the UK CMA bid are early impervious that this tightening has already started.

That favoritism matters much for SEO strategy than astir of what sewage attraction successful H1. Here is what practitioners tin do pinch it heading into the 2nd half.

First, discontinue the presumption that a azygous rank-tracking instrumentality covers AI search. Build a punctual sheet that spans ChatGPT, Claude, AI Overviews, AI Mode, and astatine slightest 1 open-weight model, and dainty the results the measurement a pollster treats a sample alternatively than the measurement an SEO treats a SERP. Given really small overlap Indig recovered betwixt engines, search only 1 is adjacent to search none.

Second, displacement reporting distant from citation counts and toward mention frequency, sentiment, and proposal rank crossed that panel. A dashboard that only counts citations is measuring the smaller half of what really drives purchaser behavior, and it will support understating your existent AI footprint to clients aliases leadership.

Third, commencement auditing your agentic entree layer now, good earlier H2 forces the issue. Check whether your merchandise data, pricing, and checkout flow are system truthful an authorized AI supplier tin really complete a transaction connected a customer’s behalf, and whether your marque shows up arsenic a trusted, nameable action erstwhile an supplier is choosing among respective competitors. Indig’s divided betwixt intelligence and agency intends the winners successful H2 will beryllium the brands agents are permitted to enactment connected behalf of, not simply the brands that get mentioned astir often.

I work together pinch Indig that the expertise to measurement AI’s effect fell down the effect itself successful H1 2026, and I do not deliberation that spread closes connected its ain betwixt now and the extremity of the year.

More Resources: