Reduce traffic drops from AI overviews

To reduce traffic drops from AI Overviews, you need to identify which searches are losing clicks, strengthen the reasons to visit your site, and build audience relationships beyond Google. Better rankings alone may not restore traffic when the search results page already answers the reader’s question.

The practical response is not to abandon SEO or rewrite every article for an AI citation. It is to protect valuable search demand, improve content that offers something beyond a summary, and measure success through qualified visits, registrations, revenue, and returning readers,not pageviews alone.

How to reduce traffic drops from AI Overviews

Identify queries losing clicks despite stable visibility, check whether AI Overviews answer them directly, and prioritize pages with meaningful business value. Maintain core SEO, add original evidence or useful tools, improve onward journeys, and develop email, direct, and other audience channels. Measure retained value as well as traffic.

This approach separates three problems that often get mixed together: losing visibility in search, losing clicks despite remaining visible, and losing business value because fewer people reach your site. Each requires a different response. A crawlability problem needs a technical fix; an answered-on-Google query needs a stronger reason to click; weak conversion needs a better reader journey.

Google Search Central says established SEO practices still matter for AI Overviews and AI Mode. Its guidance describes AI features as surfacing relevant links that help users explore content. That makes accessibility, relevance, and content quality necessary foundations, even when they cannot guarantee a referral.

The reported publisher data also makes clear why a broader strategy matters. Piano benchmark data, reported by Digiday, found organic click-through rates were 61% lower on queries where an AI Overview appeared. That is a finding from a particular dataset, not a forecast that every website will lose the same proportion of clicks.

Before changing your content strategy, define what you are trying to protect:

• Audience reach: readers who discover your publication or brand through useful content.
• Commercial outcomes: purchases, qualified leads, subscriptions, or other meaningful actions.
• Audience relationships: newsletter subscribers, registered users, and repeat visitors.
• Content economics: the return from producing and maintaining each topic or content type.

These goals can lead to different decisions. An advertising-funded publisher may need to protect pageviews and improve repeat readership. A specialist service business may accept fewer visits if the remaining visitors are better qualified. Neither should treat every lost informational click as equally valuable.

Start with the pages and query groups that connect most directly to those goals. This prevents a common mistake: spending weeks recovering visits to a generic explainer while ignoring a smaller decline on a page that regularly produces customers or subscribers.

Diagnose the decline before blaming AI search

A traffic decline that happens alongside AI search growth is not, by itself, proof that AI Overviews caused it. Changes in rankings, demand, publishing frequency, technical accessibility, and the search results layout can overlap. Your first job is to establish which explanation best fits each affected topic.

Separate visibility loss from click loss

Use Search Console to compare clicks, impressions, click-through rate, and position for affected pages and queries. Keep comparisons as consistent as possible across country, device, search type, and seasonal demand. An aggregate sitewide average can conceal very different patterns.

1. Find the affected landing pages. Identify where search visits fell and which pages account for the largest commercially meaningful losses.
2. Inspect their query groups. Separate definitions, instructions, comparisons, current events, and purchase-oriented searches rather than treating the page as one undifferentiated unit.
3. Compare visibility and clicks. Stable impressions with fewer clicks suggests a different problem from a simultaneous decline in impressions and rankings.
4. Inspect representative results. Record whether an AI Overview appears, what it answers, which sources it shows, and what competing result features occupy the page.
5. Check alternative explanations. Review indexing, redirects, site changes, content freshness, and changes in the subject’s popularity.

Manual search checks are observations, not a complete measurement system. Results can differ between users and contexts, so record the conditions of each check and avoid treating one screenshot as representative of every impression.

Use dedicated AI reporting alongside broader search data

Google’s Generative AI Performance Reports, introduced in June 2026 according to the supplied reporting, provide dedicated measurement for AI search surfaces such as AI Overviews and AI Mode. Use those reports where available, alongside broader Search Console and site analytics data, rather than expecting any one report to explain the entire decline.

The most useful working record connects a topic, its important queries, the landing page, observed search features, and business outcomes. Add an assessment of confidence: confirmed technical issue, likely search-layout impact, changing demand, or unresolved. That small distinction helps teams avoid presenting an assumption as a diagnosis.

For example, a definition page with stable visibility but fewer visits deserves investigation for answer substitution. A product guide that lost rankings after a site migration needs technical and relevance checks first. Both pages lost traffic, but applying the same remedy would waste effort.

Prioritize queries by answerability and business value

The biggest exposure is often informational content that competes directly with an answer on the results page. Search Engine Land has reported sector-level organic traffic declines ranging from 15% to 64% since AI Overviews launched, especially for informational rather than transactional content. Those observations identify a risk pattern; they do not establish a universal loss rate or isolate a single cause.

Review whether each query can be satisfied with a short explanation. A simple definition may give the searcher little reason to continue. A decision involving personal constraints, current availability, detailed evidence, or an interactive task may leave more work for the destination page to do.

Group opportunities by the action they deserve

• Protect: commercially important pages where a search visit still supports a purchase, inquiry, subscription, or other valuable action.
• Improve: useful informational pages that could offer original evidence, a better workflow, or meaningful next steps.
• Consolidate: overlapping pages that repeat the same answer without serving distinct reader needs.
• Maintain selectively: basic reference content that supports readers or the wider site but no longer justifies frequent expansion.
• Retire carefully: outdated or redundant material with no clear user purpose, after checking its remaining value and dependencies.

Do not equate low traffic with low value. A narrow technical guide may attract few readers but resolve an important buying concern. Conversely, a high-volume explainer may generate visits that rarely lead to another useful interaction.

For each topic, assess the size of the loss, the importance of the audience, the remaining reason to click, and the effort required to improve the page. Use those judgments to choose work, not to manufacture a falsely precise score. The objective is a defensible editorial decision.

Category-level findings can provide context without replacing your own data. A Similarweb-based analysis reported faster-than-average declines in health, food, news, and adult content. It also found that losses were not uniform across site types, with smaller sites often flatter overall even while publisher traffic declined.

That differs from Chartbeat findings reported by Search Engine Land, which showed particularly steep losses among smaller publishers. Different samples and definitions can produce different pictures. Use both as reasons to investigate your own exposure, not as grounds to assume that site size alone determines the outcome.

Create content that gives readers a reason to click

If your page offers only the same short answer that appears in an AI Overview, adding more words is unlikely to solve the underlying problem. The better question is what the reader still needs after receiving that answer. Build the page around that remaining task.

Useful differentiation usually comes from evidence, specificity, interaction, or access. These are editorial investments rather than formatting tricks. Choose the approach that fits your expertise and your audience instead of adding every possible feature to every page.

Add evidence that belongs to your organization

Where you genuinely have relevant experience, show it through documented testing, original photographs, interviews, worked examples, or clearly explained internal analysis. State how the evidence was gathered and where its limits lie. Do not label an article as tested or expert-reviewed unless that work actually happened.

For a software comparison, this could mean demonstrating a workflow with screenshots and explaining where the product failed a specific requirement. For a recipe, it could mean showing the visual cues that distinguish a successful result from an undercooked one. For a service business, it could mean explaining how different customer constraints change the recommendation.

Help the reader complete a task

• Provide a usable resource: a checklist, worksheet, template, or calculator that supports the next step.
• Show exceptions: explain when the standard answer does not apply and how the reader can recognize that situation.
• Make trade-offs explicit: describe what a reader gains and gives up when choosing between realistic options.
• Connect explanation to action: help readers inspect a product, compare relevant services, or work through an implementation.

These additions should solve a genuine problem. A downloadable file that merely repeats the article creates friction without adding value. An interactive tool needs clear assumptions, maintenance, and an explanation of what its output means.

Keep the basic answer easy to find. Hiding it beneath a long introduction makes the page harder to use and does not create a durable advantage. Give the concise explanation first, then earn the visit with depth, evidence, and practical help.

There is also a cost trade-off. Original research and tools require resources, so reserve the most intensive work for topics where differentiation can support a worthwhile outcome. Some reference pages should remain short, accurate, and inexpensive to maintain rather than becoming elaborate projects.

Keep core SEO strong while making content easy to understand

Answer engine optimization and generative engine optimization,often shortened to AEO and GEO,are useful labels for work aimed at AI-mediated discovery. Search Engine Land’s coverage of the Reuters Institute report describes agencies adapting existing SEO playbooks for chatbots and overview boxes. That does not mean the foundations of search optimization have stopped mattering.

Google Search Central’s guidance is the practical starting point: maintain core SEO, ensure content can be accessed, and publish useful material. A page cannot benefit from better explanation or stronger evidence if a technical mistake prevents the intended search systems from reaching it.

Audit accessibility before rewriting

• Check whether important pages are crawlable and indexable as intended.
• Review canonical signals, redirects, and internal links after site changes.
• Confirm that essential explanations and resources are available in the accessible page content.
• Check that navigation helps readers and search systems discover related material.
• Review the mobile experience so the page remains usable after a search click.

Then improve the page’s structure. Use descriptive ings, define specialist terms, and make the relationship between a claim and its supporting evidence clear. Put qualifications next to the relevant claim rather than hiding them in a distant disclaimer.

A strong explanation often follows a simple sequence: answer the question, describe the conditions under which the answer applies, show evidence or an example, and explain the next decision. That structure helps readers evaluate the content without forcing them to reconstruct your reasoning.

Make expertise and provenance visible

Use accurate bylines, relevant author information, and clear review responsibilities where appropriate. Link to the sources behind important factual claims when you have those sources available. If you publish original analysis, explain the method sufficiently for readers to understand what was measured and what was not.

For sensitive topics such as health, distinguish general information from individual advice and use qualified review when the subject requires it. Update material because something meaningful changed, not merely to display a newer date.

No ing style, markup choice, or answer-block formula guarantees an AI citation. Treat clearer structure as a way to improve comprehension and eligibility for discovery, then evaluate actual performance. Avoid buying certainty from a tactic that cannot demonstrate it.

Turn search visitors into a returning audience

Reducing the business impact of lost search traffic is partly an audience-development problem. Digiday has reported that publishers are putting more emphasis on registrations, engagement, and audience relationships as referrals become less dependable. The practical goal is to make a useful first visit capable of becoming a second visit.

That does not mean surrounding every article with competing subscription prompts. The next step should match the reason the reader arrived. Someone using a troubleshooting guide has a different immediate need from someone following an ongoing news story.

Offer a relevant next step

• For recurring questions: offer an email series or newsletter that addresses the next problems readers typically face.
• For ongoing coverage: provide a clear way to follow the topic or receive meaningful updates.
• For practical resources: explain the benefit of saving a checklist, keeping a project, or creating an account.
• For purchase research: offer an appropriate comparison, consultation, product view, or buying resource.

State what the reader will receive and why registration or subscription is useful. Ask for only the information needed for that relationship, and make the choice understandable. A registration total is not a success if the audience never returns or did not understand what it signed up for.

Internal journeys also matter. Link to the next relevant task within the article, not just to loosely related pages chosen to increase pageviews. A reader who has learned how to diagnose a problem may need a repair guide, a parts reference, or a way to assess whether professional help is appropriate.

There are trade-offs. A hard registration wall may reduce casual reading and complicate the acquisition journey. An ungated resource may reach more people but produce fewer identifiable relationships. Test these decisions against the role of the content and the value readers receive.

Measure the quality of the relationship after the initial action. Look at subsequent engagement, repeat visits, subscription retention, or qualified commercial activity as appropriate. This helps prevent a shift from chasing empty pageviews to chasing equally empty registrations.

The aim is not to eliminate search dependence overnight. It is to make each relevant visit more useful to the reader and more capable of supporting a sustainable audience relationship.

Diversify referrals without expecting AI traffic to replace Google

Fast growth from a small referral channel can sound more important than it is. Chartbeat data reported by Search Engine Land found ChatGPT referrals rose 200% while still accounting for less than 1% of total publisher traffic. In the same publisher dataset, Google Search pageviews fell 34% year over year and Google Discover fell 15%.

Those figures explain why AI discovery is worth monitoring but should not be treated as an automatic replacement for lost search visits. Growth rate and absolute contribution answer different questions. Your plan needs to account for both.

Digiday also reported Brainlabs findings of average organic traffic declines above 10% after Google’s Overview feature, while some retail, beauty, and entertainment clients saw relatively stronger AI referral gains. This suggests opportunities vary by vertical, not that every publisher can recover its losses through chatbot referrals.

Choose channels that fit how your audience behaves

• Email: use recurring editorial value to support an ongoing relationship with readers who choose to subscribe.
• Direct visits: give people a recognizable reason to return, such as a useful resource, regular briefing, or specialist coverage.
• Messaging and community: participate where the audience already exchanges relevant information, with appropriate permission and moderation.
• Partnerships: explore relevant organizations, publications, or specialists whose audiences overlap with yours.
• AI referrals: monitor the sources you can identify and evaluate the value of the visits they send.

Pick a small number of channels you can serve consistently. Maintaining a newsletter, community, video operation, and several social feeds without enough editorial capacity can weaken all of them. Distribution should extend your strongest work, not become a separate production treadmill.

Track channel mix as well as total visits. Similarweb-based analyses have described sites shifting toward direct, internal, and messaging traffic, changes that can conceal search losses in a topline audience number. Conversely, total traffic can fall while a more valuable segment remains stable.

Referral attribution has limits, particularly when people move between apps, devices, and later direct visits. Use the data you can observe, document attribution gaps, and avoid assigning every unexplained visit to AI. The purpose of measurement is to guide investment, not to produce a perfectly tidy channel story.

Measure retained value, not just recovered pageviews

Your reporting should show whether the response is protecting the business, not merely whether a chart has stopped falling. Piano’s publisher benchmarks, reported by Digiday, showed search traffic down 36%, revenue down 16%, and total audience down 9%. These are distinct measures, and their different movements illustrate why traffic alone cannot describe the full impact.

Do not combine those figures with the click-through-rate or Chartbeat findings into one implied forecast. They come from different analyses and describe different outcomes. Use them as context for selecting your own measures.

Build a scorecard around the affected topics

• Search visibility: impressions and position trends for meaningful query groups.
• Click capture: clicks and click-through rate, with relevant AI search observations or reporting.
• Visit quality: relevant onward actions and engagement with the resource the reader came to use.
• Audience development: subscriptions, registrations, and later returns attributable where possible.
• Business value: revenue, qualified leads, purchases, or another outcome appropriate to the site.
• Production cost: the effort required to create, improve, and maintain the content.

For an advertising-funded publisher, include the consequences of fewer monetizable visits. Digiday reported Ozone benchmark data showing publisher ad supply fell by up to 40% in Q2 2026, linked in that reporting to zero-click search and reduced pageviews. That is a reason to examine your own inventory and revenue exposure, not an assumption about your expected decline.

For a lead-generation site, examine whether the affected queries ever produced qualified inquiries. Losing low-intent visits may matter less than losing a small number of high-value prospects. For a subscription publisher, look at both acquisition and the longer-term value of the readers acquired.

Evaluate changes in topic-level groups rather than declaring success from a single page. Record what changed, when it changed, and what other events may have influenced the outcome. Where feasible, compare with similar pages that were not changed, while acknowledging that their search conditions may differ.

A useful result may be more qualified actions from fewer visits, stronger retention from the same audience, or a lower maintenance cost for content with reduced demand. Define that result before implementation so the team does not choose a convenient success measure afterward.

Build a focused recovery plan instead of rewriting everything

The most workable response is a sequence of diagnosis, selective improvement, and review. A sitewide rewrite makes it harder to identify what helped and can consume resources that would be better spent on a smaller number of consequential topics. Start with a manageable set of pages whose outcomes you can evaluate.

1. Establish the baseline. Export the relevant search and analytics data, document major site changes, and identify the query groups contributing most to the decline.
2. Assign a likely cause. Distinguish technical problems, lost rankings, changing demand, and reduced clicks on answerable queries. Keep unresolved cases visible.
3. Select the intervention. Choose a technical fix, content improvement, consolidation, better onward journey, or lower-cost maintenance approach according to the diagnosis.
4. Document the hypothesis. Explain what reader need the change addresses and which measurable outcome should improve.
5. Publish and distribute deliberately. Make the improvement accessible through relevant internal links and appropriate audience channels.
6. Review and adjust. Compare results with the baseline, consider other changes in the search environment, and decide whether to expand, revise, or stop the approach.

Set expectations according to the intervention. Fixing an access problem and developing an original research resource are different kinds of work. Neither justifies a promise that lost clicks will return on a fixed schedule.

Assign ownership across SEO, editorial, analytics, audience development, and commercial teams where those roles exist. SEO can identify an exposed query, but editorial may need to produce original evidence, product teams may need to build a useful tool, and audience teams may need to create the return journey.

For a small team, the same responsibilities can sit with fewer people. Keep the plan simple: one diagnosis, one primary reader improvement, one relevant next step, and a small set of outcome measures for each priority topic. Avoid collecting data that no will use to make a decision.

Finally, define what you will stop doing. That might mean producing fewer interchangeable explainers, reducing unnecessary updates to low-value pages, or declining distribution channels you cannot maintain. Protecting editorial capacity is part of adapting to a less dependable referral environment.

To reduce traffic drops from AI Overviews, focus on what you can diagnose and improve: query-level performance, technical accessibility, distinctive usefulness, and the path from a first visit to an ongoing relationship. AI citations and visibility can contribute, but they are not substitutes for reader value or measurable business outcomes.

Start by reviewing your most consequential declining query groups and choosing a focused set of pages to improve. Give each change a clear hypothesis, a reader benefit, and an outcome measure,then use the results to decide where to invest next.