If organic traffic is sliding and paid is getting pricier, the fix isn’t “replace the visits.” It’s to replace the visibility you used to get for free—before AI started answering without sending clicks.
Because the uncomfortable math is here: when Google AI Overviews show up, organic click-through rate can drop by about 70%, according to Seer Interactive findings cited by Column Five. Same ranking. Fewer visits. Less pipeline signal.
And the buyer behavior shift is no longer a niche edge case. Column Five, citing Demand Gen Report, says 25% of B2B buyers now use generative AI over traditional search for vendor research. Column Five also cites G2 survey data showing 50% of buyers start their software buying journey in an AI chatbot (up 71% in four months). That’s not a rounding error. That’s a reroute.
So the problem isn’t “traffic is down.” The problem is you’re getting less observable influence per unit of content effort. And that breaks planning.
The new constraint: influence without clicks (and without attribution comfort)
Traditional SEO KPIs assumed a clean chain: rank → click → session → conversion → pipeline. AI answer experiences cut the chain in the middle. The buyer still learns. They just don’t always arrive.
Brick Marketing’s framing is blunt: visibility is increasingly positioned as a core KPI that replaces (or at least supersedes) traffic because AI-driven answer experiences reduce the need to click through. The better read is also in Brick Marketing’s nuance: traffic still matters, but it’s incomplete on its own. Teams need a combined model—visibility and traffic.
Here’s the pattern interrupt for CMOs: a “successful” SEO program can now deliver less qualified pipeline even if rankings hold. Not because content got worse, but because distribution changed.
And there’s a second twist. AI answer engines don’t just summarize your site; they often cite third-party sources—review platforms and community/video sites in particular, per Arcade. That means the surfaces that shape preference are increasingly off-site, outside web analytics, and often outside the marketing team’s normal operating cadence.
If you only change one thing, change this: run an AI visibility holdout
The most practical move isn’t “do more content.” It’s to set up a measurement loop that tells you, quickly, whether visibility is being lost—and where you can win it back without guessing.
One primary tactic: run a visibility holdout experiment that treats AI answers and third-party citations as first-class distribution, then measures lift in qualified pipeline using directional attribution plus a clean baseline.
The hypothesis (make it falsifiable): If we publish and syndicate balanced comparison content for a defined set of buyer prompts, then our AI-answer and third-party visibility will increase and sales-qualified pipeline from that segment will rise, because answer engines disproportionately cite comparison-style sources and buyers are starting research inside chatbots.
This is grounded in two research-brief claims from Arcade: AI engines frequently cite third-party sources, and balanced comparison content (think “Tool A vs Tool B” with tables and direct answers) is more likely to be cited than product-only pages.
But the experiment needs a control. Otherwise it’s vibes.
Run it this week: setup, launch, readout, next test
Here’s the 5-minute version you can run this week:
Setup (Day 1)
- Audience: one ICP segment you can isolate in CRM (industry or firmographic slice), plus a matched “holdout” slice you won’t target with new comparison content for 4 weeks.
- Prompts to target: 10–20 evaluation queries buyers ask in AI tools (e.g., “best [category] for [use case],” “[vendor] vs [vendor],” “alternatives to [vendor]”). Keep them tied to real deal conversations and sales call notes. No guessing.
- Assets: 2 comparison pages with a table, direct answers, and clear positioning; 1 neutral explainer that defines the category and decision criteria. Arcade’s point is the key: balanced pages get cited more often than product-only pages.
- Owners: Demand gen (experiment owner), SEO/content (production), RevOps (segmentation + pipeline reporting), PR/comms (earned-media hooks). Coordination is the hard part.
Launch (Days 2–7)
- Distribution: publish on-site, then push the same decision criteria into third-party surfaces you control (review profiles, community Q&A, partner directories where appropriate). Arcade notes AI engines cite review/community/video sources—so don’t treat these as “nice to have.”
- Earned media angle: pitch one data-backed POV that maps to the comparison criteria. Corporate Ink, citing Muck Rack analysis, reports 89% of sources cited by top LLMs come from earned media. That doesn’t mean “do PR.” It means PR is now part of search distribution.
Readout (Weeks 2–4)
- Primary metric: qualified pipeline lift in the exposed segment vs holdout (directional incrementality). Use a simple diff-in-diff if RevOps can support it; otherwise, keep the baseline tight and call it directional.
- Secondary metrics: (1) AI-answer presence for the prompt set (manual spot checks on a schedule), (2) third-party mentions/citations on target sites (review platforms and relevant communities), (3) branded search impressions (not clicks) as a leading indicator.
- Stop-loss threshold: if after 4 weeks you see no movement in AI-answer presence and qualified pipeline is down >10% vs holdout, pause new production and diagnose distribution surfaces before shipping more pages.
Next test (Week 5)
- Expand from 2 to 6 comparison pages only if the visibility signals moved first. If they didn’t, you don’t have a content problem—you have a credibility/distribution problem.
The trade-off: volume will dip before quality shows up
This approach is not a traffic replacement plan. It’s a visibility replacement plan. That has a cost: it often reduces top-of-funnel volume in the short term because the work is more specific, more evaluative, and less “how-to” oriented.
But the alternative is worse: chasing lost clicks while buyers move upstream into AI tools. Orbit Media survey data cited by Column Five says 40% of users researching “how-to” topics already choose AI tools over search engines. Meanwhile, Column Five reports 47% of B2B buyers prefer ChatGPT over other LLMs. The channel mix is shifting under the dashboard.
And if the concern is, “Is this even real for SaaS?” Demand Gen Report’s DerivateX study (as cited in the brief) found 44% of B2B SaaS companies score below 50 on AI visibility, with an average AI Presence Score of 56.9/100. That’s a measurable gap, not a vibe.
When this is wrong: if the motion is high-intent, bottom-funnel search where clicks still happen (pricing, login, direct competitor queries without AI Overviews), traffic-first SEO can still carry its weight. The point isn’t to abandon visits. It’s to stop treating visits as the only proof you existed.
Traffic can fall for lots of reasons. Visibility is the one that quietly breaks the plan—because the buyer still moves, and the dashboard goes dark.