Three out of four US marketing leaders now say their core measurement approaches are underperforming, according to a 2026 IAB study reported by eMarketer. That includes attribution, incrementality testing, and marketing mix models, the entire stack most companies use to decide where their next dollar goes. Cox Automotive ran a parallel finding from the buyer side: 25% of new-vehicle shoppers used AI tools during the research phase in 2025, the first year the metric was tracked. Two surveys, two angles, one conclusion. The instruments have stopped agreeing with the room.

What Actually Changed: AI Took the Middle of the Funnel

The 2026 attribution problem is not a tracking bug. It is a missing-data problem.

C3 Metrics founder Greg Collins frames it in a way that should reorganize how marketing leaders think about their dashboards. Every customer journey has an Originator (the touchpoint that creates preference) and a Converter (the touchpoint that closes). Until 2024, both lived inside measurable channels: search results, comparison sites, third-party reviews. In 2026, the Originator has moved inside an LLM session. As Collins puts it: "The consumer was already going to buy. The brand search was how they returned to complete the purchase."

That session leaves no analytics signal. No referrer header, no UTM, no clickstream event. The journey happened. The instrumentation does not see it.

The macro evidence is now unambiguous. Chartbeat reported global organic Google search traffic dropped 33% between November 2024 and November 2025, with US traffic falling 38%. Search Engine Land found organic CTR slid from 44.2% to 40.3% over the same window. Analytic Partners reports that more than 60% of Google searches now end without an external click, and Bain found 80% of consumers rely on zero-click results for at least 40% of their searches. These are not edge-case numbers. The middle of the buyer journey moved into a private interface, and most attribution stacks were built assuming it would always be public.

For an SMB founder running marketing on top of running the business, this shows up first as a feeling. Pipeline is thinner. The dashboards still look fine. Branded search may even be holding steady. What looks like brand strength is often the residue of an AI session that nobody can see.

The Affiliate Question: Dying or Mutating?

The editorial conventional wisdom says affiliate programs are the obvious casualty of zero-click search. The numbers say something almost the opposite.

US affiliate spending will reach $13.81 billion in 2026, up 11.3% from 2025, according to eMarketer. The channel will generate roughly $241 billion in US ecommerce sales this year. On Awin's network, creator revenue share grew from 15.9% to 19.5% year-over-year. Adobe data showed a 4,700% surge in US retail traffic from generative AI sources by mid-2025.

The affiliate channel is not collapsing. It is being repositioned, and most measurement systems have not noticed.

Here is the structural reason why. McKinsey research cited by both Awin's Adam Weiss and impact.com CMO Cristy Garcia found that brand-owned content accounts for only 5% to 10% of the sources cited in AI-generated answers. The remaining 90% to 95% is third-party, affiliate publishers, review sites, user-generated content, news outlets. An Evertune analysis confirmed the same direction from a different angle: more than 40% of the top 10,000 AI-cited sources contain affiliate links or sponsored content. Acceleration Partners ran their own study and found over 80% of LLM citations originate from ad-funded websites rather than brand-owned content.

Three independent measurements, one direction. AI engines do not pull from brand sites. They pull from the partner ecosystem.

The empirical proof is already on dashboards if marketers look. eMarketer reported that nearly 70% of the sites cited in ChatGPT's mentions of eyewear brand Zenni came from affiliate marketing content. More than a quarter of OpenAI's content partnerships since 2021 have been with scaled affiliate commerce publishers.

What this means for affiliate programs is the inverse of what most operators expect. The channel that was historically credited only at the click is now structurally responsible for AI visibility, and almost nobody is measuring that contribution. eMarketer found that 27.3% of marketers fold affiliate into a generic performance bucket inside their MMM, while 14.8% do not represent it at all. The biggest growth lever in AI-era discovery sits in the spreadsheet column marked "other."

Cristy Garcia at impact.com calls this the Answer Era and frames the implication directly: "Trust is the new ranking factor. Partners are the source." Affiliate's underlying mechanism, what she calls "word of mouth with accountability built in", is what AI engines actually need to ground their answers.

The Measurement Lie Hiding in Your Dashboards

The most concrete number on dark AI traffic comes from Loamly founder Marco Di Cesare's analysis of 446,405 visits. Of all AI-driven traffic in that dataset, 70.6% landed in GA4 as "Direct" rather than being attributed to its actual AI source. Dark AI visitors converted at 4.1 times the rate of non-AI traffic in the same dataset, with a 10.21% transactional rate against a 2.46% baseline.

Three mechanisms cause the misattribution. Users frequently copy URLs from ChatGPT and paste them into a browser, which strips referrer headers. ChatGPT's mobile apps sandbox referral data entirely. Different AI platforms handle referrers inconsistently, so unified tracking is impossible without a detection layer most marketers do not have.

Stack this on top of the older problem. Braze content lead Sally Wills points out that traditional last-click attribution overstates branded search and direct traffic value by 15% to 30%. Sinuate Media's Joe Sutton notes that platforms also double-count: each ad platform claims the same conversion within its own attribution window, so a marketer adding up Google's, Meta's, and TikTok's reports often sees more conversions than they actually had.

The result is a dashboard that looks healthy and often is not. Direct traffic rises and gets read as brand strength. Branded search holds steady and gets credited to last-click. The dark AI session that originated both never appears anywhere. The publisher trade press has already absorbed the human-scale version of this story: Business Insider lost 55% of its traffic between 2022 and 2025, HuffPost is down 50%, Stereogum reports a 70% reduction in ad revenue.

For the SMB whose pipeline is softening, the operational problem is sharper. The instruments are not just imprecise. They are biased toward measuring what is easy to count, which is no longer where the demand is being formed.

The Platforms Themselves Are Diverging

The instinct to build a single, unified attribution playbook for AI search assumes the AI platforms behave like the search engines they are replacing. They do not.

Perplexity announced in early 2026 that it had permanently removed advertising from the platform after a sub-12-month experiment. The decision was strategic, not financial: Perplexity processes 780 million monthly queries and reached $200 million in annualized revenue by late 2025. A Perplexity executive explained the reasoning to Affiverse: "A user needs to believe this is the best possible answer. A user would just start doubting everything, which is why we don't see it as a fruitful thing." Subscription plus organic citation is the model. The paid path is closed.

OpenAI moved the opposite direction. ChatGPT launched advertising on free and lower-tier plans in February 2026, with sponsored results placed below organic answers. Google integrated ads into AI Overviews and AI Mode while keeping Gemini ad-free as a standalone product. Three platforms, three monetization models, three different sets of rules for marketers who want visibility.

The implication is that any unified attribution strategy will fragment as the platforms diverge. A program that depends on paid placement on Perplexity has no path forward. A program that ignores affiliate citation on ChatGPT will miss what eMarketer found in the Zenni case: 70% of brand mentions came from affiliate content. The measurement infrastructure has to treat AI organic and AI sponsored as different channels with different mechanics, not lump them together.

C3 Metrics emphasizes a related risk: when AI ad spend scales, platform self-reporting on AI advertising creates structural conflicts of interest. Independent measurement matters more in this environment, not less.

Why Better Math Will Not Save the Stack

Marketing technology vendors are converging on a common pitch: AI will rebuild attribution. The eMarketer measurement study found that 50% of buy-side marketers are already scaling AI in measurement workflows, and 69% of analytics teams are leading the adoption. There is real value in that direction. AI does compress annual MMM updates into weekly cycles. It does help cross-check outputs across methods.

It does not solve the underlying problem.

C3 Metrics is direct about this. No incremental tracking solution recovers signal that was never collected. If the Originator touchpoint happened inside an LLM with no referrer headers, no analytics provider can retroactively attribute it. Brandlight CPO Alex Prober makes the same point with different language: "The influence undoubtedly occurred, but it happened invisibly to our standard analytics." The right response is not better attribution math. It is to deprioritize attribution as the primary measurement framework and shift toward managing AI presence directly.

Analytic Partners, an MMM specialist with 25-plus years in commercial analytics, reaches a parallel conclusion from a different starting point. Their argument is methodological. The disciplines that survive when digital signals fail are older than digital, Marketing Mix Modeling and incrementality testing, because they treat search as one input among many and measure lift through controlled experiments and aggregated data, not user-level tracking.

The honest synthesis is uncomfortable for vendors and clarifying for operators. AI helps. AI does not fix. The companies that will navigate 2026 well are the ones that adopt AI in measurement while accepting that part of their demand is now genuinely unmeasurable at the user level.

What an SMB Founder Should Actually Do

Most of the published guidance assumes a measurement team exists. For a $5 million B2B services firm with no analytics function, the playbook needs to be simpler and grounded in what is already available.

Triangulate from systems that record actual revenue. A founder already knows what revenue closed, where it came from, and how long the sales cycle was. CRM, booking software, and POS systems are the source of truth, not GA4. Joe Sutton at Sinuate Media recommends Marketing Efficiency Ratio (MER) at the macro level: total revenue divided by total marketing spend, tracked over rolling windows. MER does not care which channel converted. It tells the founder whether the marketing budget is producing pipeline at all.

Audit the AI presence that already exists. Spend an afternoon asking ChatGPT, Perplexity, Gemini, and Claude the queries a buyer would actually type, "best [category] for [use case]," "alternatives to [competitor]," "how does [your category] handle [problem]." Note which sources the AI cites in its answers. If it cites competitors and ignores the company, that is the new SEO problem. If it cites the company but cites a partner more often, that is the new affiliate problem.

Treat partner relationships as discovery infrastructure. The eMarketer FAQ recommends moving past last-click in affiliate payouts toward multi-touch or data-driven models. For SMBs, the version of that recommendation is to stop evaluating partners only on the conversions their UTM links closed. Evaluate them also on whether they get cited by AI tools when buyers ask category questions. Acceleration Partners argues for "influence-based frameworks" that reward affiliates for being trusted sources, not just last-touch closers.

Run small experiments instead of trying to track everything. Hold out a market for 30 days. Pause one channel for two weeks. Compare. Incrementality at the SMB scale is two regions and a calendar, not a Bayesian model. The data is noisier than enterprise teams would tolerate, but the answers are more honest than the dashboards.

Stop chasing attribution. Start managing presence. Brandlight's Alex Prober uses that phrase precisely because it inverts the marketing instinct of the last decade. The instinct was to instrument every touchpoint and follow the user through the funnel. The discipline that 2026 is rewarding is to make sure the brand is correctly represented in AI answers, with consistent narrative across platforms, before someone asks about it.

The Deeper Shift

The attribution model breaking is not a measurement story. It is a story about who controls the middle of the buyer journey.

For two decades, that middle belonged to platforms marketers could buy access to. Search engines, comparison sites, review aggregators, social feeds, all measurable, all biddable, all instrumented. In 2026, a meaningful share of that middle has moved into private LLM sessions where no instrument reaches. The question for any company growing through 2026 and 2027 is not "which attribution model should I switch to." It is "where is my brand showing up in AI answers, and which partners are putting it there."

The companies that figure that out first will not have better dashboards than their competitors. They will have a more honest one.

Sources