Visibility Now Happens Before the Click, and Attribution Is Breaking

Picture a quarterly review at a B2B SaaS company. The organic traffic chart slopes down, and someone finally asks the question everyone has been dreading: is content even working? Pipeline, oddly, looks fine. And in the last few discovery calls, a prospect mentioned they had “asked ChatGPT” before booking the demo.

Both things are true at once. That’s the whole story, and it’s why so many marketing dashboards feel wrong right now.

Most of the influence on a B2B software purchase now happens in places your analytics can’t see. A buyer reads an AI answer, skims a Reddit thread, notices a peer’s LinkedIn post, and only then types your name into a browser. By the time a visit gets recorded, the shortlist is already partly formed, and the visit often lands in the wrong column of your report.

This piece walks through the strongest 2026 evidence on zero-click search, AI-assisted buying and attribution, is honest about where the data is solid and where it’s still a hunch, and ends with a measurement approach you can use. It’s written for CMOs, marketing leaders and founders who have to defend a 2027 budget with imperfect numbers.

Fewer searches end in a click

You’ve probably come across the zero-click trend already, but the latest numbers deserve a second look. SparkToro and Similarweb found that 68.01% of U.S. Google searches in the first four months of 2026 ended without a click. That’s up from 60.45% in 2024 and roughly 45% a decade ago. The count includes clicks to Google’s own properties and to ads, so it’s really measuring how often Google keeps the searcher to itself.

Two things are pushing it along. AI Overviews now appear on more than a fifth of searches, according to Ahrefs figures SparkToro cites, and where they show up they cut click-through by close to 60%. People are also running more follow-up searches without leaving Google, up about seven points between the two studies. Meanwhile, AI tools send out less than 1% of the traffic they handle.

The caveats matter. The 2024 and 2026 figures come from different clickstream panels, so read the comparison as directional. SparkToro says as much. AI Mode accounted for only 0.34% of searches in that window, even though Google says it has passed a billion monthly users. And the mobile data leaves out the Google app, where zero-click behavior is probably higher still.

This doesn’t necessarily mean demand is collapsing. SparkToro’s own view is that traffic can fall while revenue holds, and that branded, high-intent and tactical searches still reward good SEO. What has changed is that the top of your funnel increasingly lives on surfaces that don’t send visits.

Buyers decide before they visit

B2B buyers are using AI all the way through a purchase, and they tend to arrive with a short list already in their heads. Forrester’s 2026 survey of nearly 18,000 business buyers reportedly found that 94% used generative AI in their most recent purchase. G2’s March 2026 survey of 1,076 software buyers reportedly found that 51% start with an AI chatbot more often than with Google, up from 29% eleven months earlier. Both figures reached us through secondary coverage, so check the original reports before you quote them.

Using AI isn’t the same as trusting it, though. TrustRadius’s 2026 report, drawing on 1,862 technology buyers and 444 vendors, found that 63% used AI during their purchase and that 94% of those buyers fact-check the answers at least some of the time. It also found that 83% shortlisted three or fewer products. HG Insights’ reading of the same data adds that 79% had already heard of the product before formal research began.

Alexander Kesler, writing in Search Engine Journal, gives the period before formal research a name: Day Zero. Drawing on 6sense data, he argues that buyers choose a vendor that was already on their shortlist 95% of the time, and that around nine stakeholders each form a view privately before the group ever meets. He also points to Forrester data suggesting that when buyers doubt an AI answer, they drift back toward vendors they already recognize.

The part that often gets overlooked is what this says about AI itself. AI answers don’t create preference out of thin air. They reflect brand familiarity and third-party presence. So the work that earns you a place on the shortlist happens earlier, and in more places, than any visit your analytics can count.

Where the influence ends up

When AI or community influence works, it tends to show up as direct traffic and branded search, not as a tidy referral. A Similarweb study with SparkToro looked at brands recommended in AI answers across finance, travel and beauty. Over the following seven days, people who had seen a recommendation were more likely to visit that brand’s site directly. In the examples SparkToro highlights, Capital One saw a 14.2% lift and American Express 7.2%, while search visits for influenced users ran about 15% lower and direct visits rose.

A word of caution. This is consumer data on big brands, and SparkToro lists the open questions itself. Would these people have found the brands anyway? Does it hold for small or unknown brands? Nobody has published an equivalent B2B SaaS dataset yet. The mechanism is plausible for software buyers, but treat it as a hypothesis to test in your own numbers.

Your tools make this harder to see. Kevin Indig, citing Graphite, estimates that AI influence can be under-attributed by as much as ten times. That’s an estimate, not a measurement, but it points the same way as everything above. Google Analytics 4 added an AI Assistant channel in May 2026, yet Search Engine Journal notes that visits from AI Overviews still count as organic search. Search Console’s AI report has no query dimension, and a brand mention without a link doesn’t count as an impression.

The underlying data is also wobbly right now. Search Engine Roundtable reports that third-party Google rank tracking is struggling as Google blocks more scrapers, and that Search Console’s performance report ran more than 64 hours late after the September spam update.

Why traffic reports now mislead budgets

If traffic is your headline KPI, you’ll tend to cut the work that creates demand and fund the work that captures it. Kevin Indig’s 2027 planning guidance makes the case plainly. Click-based ROI has decayed, AI influence is undercounted, and a plan built only on cleanly attributable numbers will pull money away from channels that work. He quotes Ramp’s George Bonaci saying that attribution alone would have led the company to cancel brand marketing, stunts, direct mail and events.

You can spot three failure modes in most budget conversations:

• The traffic chart reads as channel failure. A falling line gets blamed on SEO or content, even when branded demand and pipeline are steady.
• Budget drifts to paid. Paid is the one channel with clean attribution, so it wins the argument by default. Indig warns that teams without a defensible AI search investment case risk losing budget to paid.
• Teams report the wrong AI metric. Rand Fishkin’s SparkToro research found that AI tools give highly inconsistent brand recommendations, so a single “rank” inside an AI answer isn’t stable. How often your brand appears across a consistent set of prompts is the more honest measure.

That doesn’t make attribution useless. It makes it one lens among several, which is exactly why a 2027 measurement plan needs more than one.

How we’d measure influence instead

We’d measure influence as a set of three signals with different blind spots, instead of trusting any single number. The idea of triangulating exposure, behavior and outcome comes from Indig’s planning memo. The B2B SaaS signals below are our own adaptation.

Report all three layers on one page every month. The value is in how they move together. Confidence rises when they agree, and a divergence tells you where to dig. Indig’s examples translate into a handy set of diagnostics:

• Exposure up, branded search flat: mentions aren’t landing with buyers. Treat it as a distribution and message problem.
• Branded search up, pipeline flat: interest isn’t converting. Fix landing pages, the offer and the handoff to sales.
• Branded and direct visits up, AI citations flat: something off-site is working and AI hasn’t caught up. Find the source, such as a podcast, event or partner, and fund more of it.
• Exposure and branded search both down: a competitor is taking the mentions. Audit who AI recommends in your category and which sources it cites for them.

No single pattern proves anything on its own. Together, though, they give you a story a CFO can follow, which a traffic line alone no longer does.

What to do in the next 90 days

Start with changes that are cheap and produce evidence you can show in the next planning cycle.

1. Baseline your AI visibility. Pick 25 to 40 prompts a buyer would actually type: category questions, comparisons and problem-led queries. Run them monthly across ChatGPT, Gemini, Perplexity and Google AI Mode, and note whether you appear, what’s said and which sources get cited. Report share of appearances, not rank. Fix wrong descriptions first, because they steer buyers away.
2. Capture self-reported attribution. Add an open-text “how did you hear about us” field to demo and trial forms, mirror it in your CRM, and ask sales to log the first-touch story from discovery calls. Free text catches influence that no dropdown anticipates.
3. Set up an AI view in analytics. Use the GA4 AI Assistant channel, annotate changes, and watch direct visits to pricing and demo pages. Keep in mind it will undercount, since AI Overview visits still count as organic search.
4. Track branded demand. Use the Search Console branded queries filter. It’s Google’s definition of branded and you can’t edit it, so trend it against your exposure signals rather than reading it alone.
5. Budget off-site presence as named hours. Indig suggests treating community and distribution as a line item, roughly 5% to 10% of marketing budget in his view, with an owner attached. A community line with no owner tends to be the first thing cut.
6. Re-split the budget every quarter. Indig’s structure works well: proven work at full capacity, learning work with one owner and a kill date, and unmeasurable bets such as community, events and brand funded on judgment.
7. Keep your own content accurate. AI answers and Google’s zero-click features still draw on your site, so pricing, product and capability pages need to describe you clearly and consistently.

What not to do

Some tempting responses make the problem worse.

• Don’t abandon SEO. Fishkin’s view is that SEO matters as much as ever, even if it earns less traffic. Lily Ray’s analysis of 11 sites that lost Google visibility found that all 11 also lost AI search citations, averaging a 22.5% drop. AI answers lean on the same ranking signals.
• Don’t present consumer AI-lift data as B2B proof. The Similarweb study covers finance, travel and beauty. Use it to justify a test, not to promise a result.
• Don’t buy or sell precise AI rankings. Given how inconsistent AI recommendations are, a confident rank number is more noise than signal.
• Don’t manufacture signals. Google has warned against fake author information, and Anthropic’s updated usage policy names seeding search and AI answer sources with misleading content. Real expert profiles, genuine community participation and earned mentions are the only durable route.

BuzzIQ’s take

The question to retire is “how much traffic did SEO drive last quarter?” The better one is “was our brand already in the room when the shortlist formed?” Traffic still matters for branded, high-intent and tactical queries, but it’s increasingly a lagging indicator of influence that happened somewhere else.

For B2B SaaS teams, that changes the job in three ways. Content has to earn mentions and citations beyond your own site, which means original data and named expertise that other sources actually want to reference. Visibility has to be planned surface by surface, because the overlap between AI engines is small and the signals that matter sit off-site. And measurement has to move from one attributable number to a monthly view of exposure, behavior and outcome that finance can follow.

We’d also keep expectations honest. The evidence for B2B-specific AI lift is still thin, the tools are unstable, and some of the loudest claims in this space rest on weak methods. That doesn’t mean you should wait. A team that tests carefully, writes down what it learns and puts kill dates on its experiments will be in a stronger spot than one that bets the budget on a single metric.

Where this applies at BuzzIQ. Organic and AI visibility work (content, SEO, AEO and GEO) covers the exposure layer. Demand generation and full-funnel programs cover the behavior and outcome layers, including how pipeline reporting is set up so it reflects influence and not only last click. If you’re planning 2027 budgets and want a baseline of how AI describes your company today, that’s a sensible first step.

Sources

• SparkToro: In 2026, less than one third of Google searches still send a click
• SparkToro: How AI brand mentions influence direct visits (Similarweb study)
• Search Engine Land: How to plan 2027 when traffic is down and AI won’t attribute (Kevin Indig)
• Search Engine Journal: 3 steps to win B2B brand consideration (Alexander Kesler)
• Search Engine Journal: AI Overviews have spread to most big-brand searches
• Lily Ray: Are citations in AI search affected by Google organic visibility changes?
• SparkToro: AIs are highly inconsistent when recommending brands
• Forrester and G2 buyer figures, as reported by Abmatic
• TrustRadius 2026 B2B Buying Disconnect Report, via HG Insights
• Search Engine Roundtable

Several figures above (Forrester, G2, 6sense, AirOps and the Graphite estimate) were cited in secondary coverage and not checked at the original source. Verify them before quoting.