Somewhere in your analytics right now is a chart that looks like a problem: organic sessions trending down, quarter after quarter, while your content output and rankings have barely moved.
Before anyone panics, it’s worth being precise about what that chart is actually telling you. In most cases it isn’t saying demand dropped. It’s saying the click went away.
AI Overviews answer the question on the results page. ChatGPT, Perplexity, and Gemini have become discovery surfaces in their own right. A growing share of searches (already the majority in many categories) end without a single click to any website. The information still gets consumed. Your brand may still get mentioned. But the session never happens, so your analytics never sees it.
This is the biggest structural change to organic marketing since mobile, and most teams are still reporting on it with a framework built for a world where the click was the whole point.
What Actually Changed
Three shifts stacked on top of each other:
- Search engines became answer engines. Google’s AI Overviews sit above the organic results and resolve informational queries in place. The click-through rate on informational content has fallen off a cliff, even for pages that still rank first.
- Assistants became discovery. People ask ChatGPT which CRM to buy, how to fix their attribution, what running shoes to get. Those conversations shape purchase decisions, and they happen entirely outside your analytics.
- The remaining clicks concentrated. Transactional and high-intent queries still send traffic. Informational queries mostly don’t. So your traffic mix is shifting toward the bottom of the funnel even if your content strategy hasn’t changed at all.
None of this means organic is dead. It means organic’s output changed from sessions to influence, and influence is harder to see in a dashboard.
The Measurement Problem Comes First
Here’s the trap: if you keep judging organic by sessions and last-click conversions, the numbers will keep getting worse, and eventually someone in finance will suggest cutting the investment right when brand presence in AI answers is becoming the thing that matters.
This is the same class of problem I wrote about in Proving Incrementality Without Perfect Attribution: the dashboard describes what it can see, not what’s happening. You need proxies for the part it can’t see.
What I’d track instead:
- AI referral traffic. Segment out sessions from chatgpt.com, perplexity.ai, gemini.google.com and similar referrers in GA4. It undercounts badly (most AI exposure never produces a click), but the trend line tells you whether assistants are surfacing you at all.
- Branded search volume. If people discover you inside an AI answer, the next observable action is often a branded search. Watch it in Search Console and Google Trends as a demand proxy.
- Direct traffic quality. Some of your “direct” is really AI-influenced discovery. Rising direct traffic to deep pages (not the homepage) is a signal worth reading.
- Share of answer. Run your priority prompts (“best X for Y”, “how do I solve Z”) through the major assistants monthly and log whether you’re cited, mentioned, or absent. Do it manually with a spreadsheet before paying a tool to do it. You’ll learn more from twenty prompts you chose than from a vendor’s dashboard of a thousand you didn’t.
- Impressions vs. clicks in Search Console. Impressions holding while clicks fall is the zero-click signature. That’s the chart to show leadership when they ask why sessions are down, because it separates “we lost visibility” from “the click disappeared.”
If your definitions and channel groupings are a mess, none of this works. That layer still comes first (the measurement stack blueprint covers it).
How to Actually Show Up in AI Answers
The honest version: nobody fully controls this, and anyone selling guaranteed AI visibility is selling confidence, not a mechanism. But the patterns in what gets cited are consistent enough to act on.
Be the source, not the summary. AI systems synthesize. Content that is itself a synthesis of other people’s information is exactly what they replace. Original data, proprietary benchmarks, real numbers from your own work, named frameworks, actual experience: that’s what earns citations, because the model has nowhere else to get it.
Win the entity, not just the keyword. Assistants recommend brands they’ve seen associated with a problem across many sources. Mentions in industry publications, reviews, comparison posts, podcasts, and communities all feed that association. This is why digital PR quietly became an AI visibility strategy, a point I made in Improving Digital Marketing and PR Cooperation years before it was about training data.
Structure content so machines can lift from it. Clear headings that match real questions, direct answers in the first sentence under each heading, tables and lists for comparisons, schema markup where it fits. You’re writing for a reader and a parser at the same time.
Stay fresh where it counts. Assistants with browsing favor current sources for time-sensitive queries. A maintained, updated page beats a new thin one.
Keep earning the clicks that still exist. Tools, calculators, templates, pricing pages, deep operator guides: things an AI answer can’t substitute for. If a page’s entire value can be restated in three sentences, an AI will restate it in three sentences.
Reframing Organic’s Job for the CFO
The conversation you want to have before the traffic chart forces it:
Organic’s job used to be delivering sessions that converted, and we measured it like a channel. Its job now is making sure that when a buyer asks a machine for a recommendation, we’re in the answer. That’s closer to brand than to performance, and it needs brand-style measurement: share of answer, branded demand, and influence proxies, with sessions as one input rather than the scoreboard.
Agree on the new KPIs while things are calm. The teams that get in trouble here aren’t the ones losing clicks (everyone is losing clicks). They’re the ones still defending a sessions target they can’t hit for structural reasons.
A 30-Day Starting Plan
Week 1: Baseline. Segment AI referrals in GA4. Pull 12 months of Search Console impressions vs. clicks. Chart branded search volume.
Week 2: Share of answer. Write down the 20 prompts that matter most for your business. Run them through Google’s AI results, ChatGPT, and Perplexity. Log citations and mentions. Calendar it monthly.
Week 3: Content triage. Sort your organic pages into three buckets: still earns clicks (protect and update), informational and being absorbed (restructure for citation, stop measuring by traffic), or neither (candidates to consolidate).
Week 4: The reframe. Take the impressions-vs-clicks chart and the share-of-answer baseline to whoever owns the budget. Propose the new KPI set before next quarter’s numbers propose it for you.
Closing Thought
Every marketing channel eventually stops paying for the behavior it was built on. Organic search paid in clicks for twenty years, long enough that whole teams forgot the clicks were the means and not the end.
The demand is still there. The buyers are still asking questions. They’re just asking a machine, and the machine is answering with whatever sources earned its trust. Your job is to be one of them, and to build the measurement that proves you are, because the session-based version of proof is going away whether you’re ready or not.