AI is eating clicks. Here is the visibility strategy that still works
The short answer
Three things are true at once, and most strategies fail because they act on only one.
Clicks are structurally down. The traffic that still arrives is better than what it replaced. And the volume of that better traffic is small today.
Act on the first alone and you cut the budget. Act on the second alone and you chase a conversion headline nobody has measured properly. Published multiples for AI traffic run from 3x to 23x. That is not a finding. It is a warning that the measurement is immature.
Key takeaways
- The decline is real and structural. With an AI Overview present, the top-ranking page earns 58 percent fewer clicks than it otherwise would.
- Citation is worth money. Cited pages earn 2.07 percent click-through against 0.94 percent uncited. Both trail the 3.35 percent of a page with no Overview.
- The arriving traffic is genuinely better. Adobe measured AI-referred retail visitors converting about 42 percent better, with revenue per visit up 37 percent.
- But nobody agrees by how much. Claimed premiums span 3x to 23x across published studies. AI referrals also remain roughly 1 percent of referral traffic for most sites.
- So change what you measure before you change what you publish. Clicks stop being the unit. Named, cited and described correctly become the units.
Where this comes from. We run organic for more than 250 enterprises at Pepper and track over 10 million prompts across every major engine. We watch this decline on live accounts weekly, and we have had the budget conversation many times. That work shapes the view here, along with the client reviews we sit in and the conversations we have at the events we run.
What is an AI search visibility strategy?
An AI search visibility strategy is a plan to be present in answers rather than to win clicks. It is measured on whether engines name and cite you, rather than on sessions.
Three things separate it from an SEO plan.
- The outcome is a mention rather than a visit. You can influence a purchase without ever appearing in analytics. That breaks most existing reporting.
- The competition is invisible. You never see the other candidates an engine considered. So competitive analysis moves from rank tracking to reading answers.
- The payoff compounds slowly. Most levers here are earned rather than bought. Quarterly reporting makes it look like nothing is happening.
That last point makes this a strategy question rather than a tactics one. If you cannot defend the reporting, the work gets cut before it matures.
See where you show up. Pepper’s GEO platform tracks Brand Visibility, Domain Prompt Presence and Share of Voice across ChatGPT, Perplexity, Gemini, Claude and Google AI Overviews. Your team can log in, connect Search Console and GA4, manage the prompt set and run your own agents in the Agent Atlas. A growth team works the same account alongside you. Book a growth audit or see where you show up.
The three facts you have to hold together
Most advice on this topic picks one and builds a strategy on it. All three are true. The strategy only works if you hold all three.
Hold them together and the conclusion is unglamorous. Do not chase the multiple. Do not cut the channel. Change the measurement, then fund the work that earns citations.
What the click decline actually looks like
The decline is well measured, and the numbers are consistent across two independent methods.
Ahrefs compared 150,000 keywords carrying AI Overviews against 150,000 informational controls. The top-ranking page earns 58 percent fewer clicks when one is present.
Seer Interactive’s 2025 averages break it down by what the page contained.
- No AI Overview: 3.35 percent organic click-through.
- AI Overview present, and you are cited: 2.07 percent.
- AI Overview present, and you are not cited: 0.94 percent.
Two readings matter. Citation earns roughly 120 percent more clicks per impression than no citation, so it is worth real money. A cited result still trails a no-Overview page by about 38 percent, so it does not restore what the format took.
We unpack the diagnostic side, including how to tell a format change from a ranking loss, in rankings held and traffic fell.
The traffic that still arrives is better
This is the half most decline commentary leaves out.
Adobe Analytics measured across more than one trillion visits to US retail sites. AI-referred visitors converted about 42 percent better than other traffic. Revenue per visit rose 37 percent, time on site 53 percent and pages per visit 23 percent.
That is a large sample from a named analytics provider, and it is the most defensible figure we have found on this question.
Now the caution, which most articles skip
Now the caution, which most articles skip. Other published conversion multiples for AI traffic range from roughly 3x to 23x. Those are not compatible with each other, nor with Adobe’s 42 percent. They measure different things. Different industries, different attribution windows, different definitions of AI traffic, and several come from vendors with a product to sell.
When published estimates of one quantity span nearly an order of magnitude, the honest conclusion is that the measurement is immature. We use Adobe’s figure for its sample size and provenance, and we do not repeat the larger multiples.
The volume caution matters more. AI referrals still account for roughly 1 percent of referral traffic for most sites. A large multiple on a small base is still a small absolute number today. The case for acting now is that the base is growing, not that it is already material.
So what actually works
Four moves, and only the last one is content production.
Work down the list rather than starting at the bottom. Most teams do the reverse. That is why their reporting never improves even when their publishing does.
The four that matter
- Change the reporting first. If clicks remain the only metric, every other change here will look like a failure. Track named, cited and described correctly, per engine, before you change any tactics. Brand visibility explains what each number means.
- Protect the retrieval floor. Verify crawler access by name, and confirm pages render without JavaScript. Free, and it gates everything. The method is in can AI search bots crawl my website.
- Work the sources that get cited instead of you. Read the source lists on answers you are missing from. That is your earned-media target list, and our piece on third-party sources covers what to do with it.
- Make the pages you already have quotable. Retitle as questions, answer in the first two sentences, and add something a model cannot compose. The evidence is in what AI engines actually cite. Updating beats publishing, since 75 percent of pages LLMs cite had been updated in the last year while only 42 percent were recently published.
Our methodology: how we weighted the response
| Lever | Weight | Why it carries this much |
|---|---|---|
| Reporting that separates named from cited | 30% | Without it the rest gets cut before it matures, which is the usual failure |
| Earned third-party coverage | 30% | 84 percent of AI citations come from earned media, against 0.3 percent from paid |
| Citability of pages you already own | 25% | Cited pages earn roughly 120 percent more clicks per impression than uncited ones |
| Retrieval floor | 15% | A gate rather than a lever, but nothing above it counts until it passes |
Strategy options at a glance
| Response | What it assumes | Typical 2026 cost | Where it works | Where it falls short |
|---|---|---|---|---|
| Cut organic investment | The decline is permanent and unrecoverable | Saves budget now | Genuinely low-exposure query sets | Gives up the compounding half, and the base is growing |
| Publish more | Volume restores the losses | The largest content line item | Almost nowhere | Freshness comes from updating, not publishing |
| Chase AI referral traffic | The conversion multiple is real and large | Varies | Retail, where Adobe’s data is strongest | The multiple is unsettled and the volume is about 1 percent |
| Buy a visibility dashboard | Seeing the problem addresses it | Roughly $99 to $400 a month | Teams with capacity to act on it | Diagnoses, never fixes |
| Change the measurement, then earn citations | Presence in answers is the unit | Reporting change is free, earned work is slow | Most B2B and considered purchases | Reports badly for two quarters |
| Platform plus a growth team | You need the work done, not described | Custom | Teams without internal capacity | Still gated by how fast others publish about you |
—
How to choose your response
The honest answer depends on your own exposure, which takes an afternoon to measure.
The scorecard
Score your position out of 100.
| Factor | Weight | How to score it honestly |
|---|---|---|
| You measured your own AI exposure | 30% | Someone ran your top queries and counted how many returned an AI answer |
| Reporting separates named from cited | 25% | Two numbers per engine appear in the monthly report, rather than one blended score |
| Third-party mention composition | 20% | Credible independent sources describe you, and you can name three without checking |
| Existing pages are quotable | 15% | Titles read as questions, and answers sit in the first two sentences |
| Retrieval floor verified | 10% | Crawlers checked individually by name, and pages render without JavaScript |
### The live test, over 90 days
Fix 30 prompts in your buyers’ words and hold them steady. Real questions, such as “how do I stop losing clicks to AI”, “best data observability platform for a small team”, “how much does sales enablement software cost” and “which GEO agency works with fintech”.
Run them across Google, ChatGPT and Perplexity. Record four columns. Whether an AI answer appeared, whether you were named, whether a page of yours was cited, and which sources appeared instead.
Rerun the identical set monthly for 90 days. Three readings is the shortest honest window. One is a snapshot, two could be noise, three shows direction. Keep the set fixed, because changing it resets the comparison.
Then put those four columns in the board deck alongside sessions. That single reporting change buys the work enough time to compound, and it costs nothing. The free method is in how to measure AI search visibility without expensive tools.
Weak approach versus strong approach
The weaker approach: watch sessions fall, publish more to compensate, then report traffic monthly. Volume does not address a format change, and the reporting guarantees the work looks like a failure.
The stronger approach: add named and cited to the report first, then protect the retrieval floor. Spend the content budget on updating existing pages and earning third-party coverage.
The weak version treats a structural change as a performance problem. The strong version accepts fewer, better visits, and measures what now drives them.
Red flags
- A conversion multiple quoted without a method. Published figures span roughly 3x to 23x, so any single one presented as settled is a sales aid.
- Traffic recovery promised. A cited page still trails a no-Overview page by about 38 percent. The old number is not coming back on those queries.
- More publishing as the answer. Updating existing pages has better evidence behind it.
- Clicks as the only metric. It guarantees that citation work reports as a failure for two quarters.
- Cutting organic entirely. The arriving traffic is measurably better and the base is growing.
- Guaranteed citations. Nobody controls generated output, and Google warns against providers guaranteeing rankings.
Five questions worth asking
- “What is our measured AI exposure?” Not the industry average, your own queries.
- “Does our reporting separate named from cited?” If not, fix that before anything else.
- “Which sources appear instead of us?” That list is the plan.
- “What is our AI referral volume today?” Usually about 1 percent, and worth knowing before anyone builds a case on it.
- “What would you tell us to stop doing?” Usually publishing more of what is not being cited.
Reduced to one principle: change the measurement before you change the tactics.
One closing note that costs us something. The first and most valuable move here is a reporting change you can make yourself in an afternoon, with no agency and no software, ours included in the general point that vendors would rather sell the dashboard than the discipline. Do that first, and judge everything after it against the new baseline.
What does this cost?
| What you are buying | Typical 2026 cost | What it covers |
|---|---|---|
| Measuring your own exposure | Free, an afternoon | Thirty queries, three engines, four columns |
| Changing the report | Free | Named and cited alongside sessions |
| Crawler and rendering checks | Free, ten minutes | The retrieval floor |
| Updating existing pages | Cheaper than new production | Where the recency evidence points |
| Earned third-party work | The largest line item, and slowest | Where 84 percent of citations originate |
| Multi-engine tracking | Roughly $99 to $400 a month | Automation once the manual version stops fitting |
| Platform plus a growth team | Custom | Pepper: tracking, plus the people and agents doing the work |
The first three rows are free and they are the ones that decide whether the rest survives a budget review.
What nobody should promise you
Traffic recovery on queries where AI answers appear. A cited page still trails a no-Overview page by roughly 38 percent.
A reliable conversion multiple for AI traffic. Published figures span roughly 3x to 23x, which means the quantity is not settled.
That AI referral volume is already material. It is around 1 percent of referral traffic for most sites, and the argument for acting is growth rather than current scale.
A guaranteed citation. Nobody controls generated output, and Google warns against providers guaranteeing rankings.
When to do nothing yet
The answer that costs us the sale. If you measured your own queries and almost none returned an AI answer, you do not need a new strategy and we would tell you not to buy one.
That is a real situation on narrow technical, transactional and local query sets. Keep doing what works, and recheck in six months. On one measured sample, AI Overview share moved from 8.7 percent to 19.0 percent in eight days around a single core update. Low exposure today is not permanent.
Frequently asked questions
Is AI really reducing search clicks?
Yes, and it is well measured. Comparing 150,000 keywords carrying AI Overviews against 150,000 controls, the top-ranking page earned 58 percent fewer clicks when an AI Overview was present.
Does being cited in an AI answer help?
Substantially. Cited pages earned 2.07 percent click-through against 0.94 percent for uncited ones, roughly 120 percent more clicks per impression. Both still trail the 3.35 percent of a page with no AI Overview.
Does AI referral traffic convert better?
The best-sourced evidence says yes. Adobe measured AI-referred retail visitors converting about 42 percent better, with revenue per visit up 37 percent, across more than a trillion visits. Larger multiples circulate, but they are not consistent with each other.
How much traffic comes from AI search?
Roughly 1 percent of referral traffic for most sites. The case for investing rests on growth and on influence you cannot see in analytics, rather than on current volume.
Should I cut my organic budget?
Rarely, and not because of this. The arriving traffic is measurably better and the base is growing. Classic ranking also still gates AI citation, since 88.46 percent of ChatGPT citations came from pages retrieved through search.
What should I measure instead of clicks?
Three things per engine: whether you are named, whether a page of yours is cited, and whether you are described correctly. Add which sources appeared instead of you, because that becomes your target list.
Will publishing more content fix the decline?
No, on the available evidence. Among pages LLMs cite, 75 percent had been updated in the last year, while only 42 percent had been published recently. Updating existing pages is the better use of the same budget.
How long before this work shows results?
Plan on three monthly readings of a fixed prompt set, so roughly 90 days for direction. The earned third-party half runs slower still, because it depends on other people publishing.
Where to go next
Before changing any tactics, add two columns to your monthly report: whether engines named you, and whether they cited a page you own.
That costs nothing, takes an afternoon to populate by hand, and it is the change that stops the rest of the work being cut before it matures.
Diagnose the decline · Learn AI Search · Book a growth audit
Sources and further reading
Studies and primary sources
- Ahrefs. Comparison of 150,000 keywords carrying AI Overviews against 150,000 informational controls, finding the top-ranking page earns 58 percent fewer clicks when an AI Overview is present.
- Seer Interactive. Organic click-through by AI Overview state, 2025 averages: 3.35 percent with no AI Overview, 2.07 percent when present and citing, 0.94 percent when present and not citing. Citation delivers roughly 120 percent more clicks per impression while still trailing a no-Overview result by about 38 percent.
- Adobe Analytics. Measurement across more than one trillion visits to US retail sites. AI-referred visitors converted about 42 percent better than other traffic, with revenue per visit up 37 percent, time on site up 53 percent and pages per visit up 23 percent.
- Muck Rack. “What Is AI Reading?” May 2026. More than 25 million cited links across ChatGPT, Claude and Gemini. 84 percent of citations from earned media, 0.3 percent from paid. Link
- Seer Interactive. Content recency study, 7,683 dated pages and 47,097 citations, March to June 2026. 75 percent of cited pages updated within a year against 42 percent published recently. Link
- Ahrefs. Study of 1.4 million ChatGPT prompts. 88.46 percent of citations came from pages retrieved through the search channel. Link
- Google Search Central. AI search optimisation guidance, 15 May 2026. Warns against providers guaranteeing rankings.
Further reading
- Pepper. Rankings held and traffic fell for the diagnostic, and earned authority for the third-party half.
- Deliberately excluded: every AI traffic conversion multiple other than Adobe’s, since published figures span roughly 3x to 23x and are not compatible with one another. Also excluded, the claim that around 68 percent of US searches end without a click, and the claim that cited brands earn 35 percent more organic click-through. Neither traces to a published method we could open.
Latest Blogs
Most advice on this is inference dressed as fact. OpenAI documents exactly one thing clearly, which is access. Here is what is documented, what is inferred, and how to verify the rest for yourself.
Being named in an AI answer and having your page used as its source are different outcomes with different causes and different fixes. The gap between them is the most useful diagnostic in AI search, and most reporting collapses it into one number.
Every platform reports share of voice and almost none of them define it the same way. Here is the definition worth using, the three choices that change the number, and how to measure it yourself without buying anything.
Get your hands on the latest news!
