Forecasting Traffic Loss From AI Overviews

Forecast traffic loss from AI Overviews by splitting your Search Console queries into exposure groups, measuring how CTR has already changed at stable position in each group, and applying low, middle and high CTR scenarios to the clicks you expect next year. Our view: base scenarios on your own data, not a headline industry figure, and present a range with stated assumptions.

Points marked "Our view" are interpretation; numbers in examples are illustrative.

On this page

• Why forecast at all?
• What data do you need?
• Step 1: How do you group queries by exposure?
• Step 2: How do you measure CTR change?
• Step 3: How do you build scenarios?
• Step 4: How do you turn CTR into clicks?
• Step 5: How do you translate to business impact?
• How should you present it?
• What actions reduce the risk?
• Which pitfalls should you watch for?
• How often should you update it?
• How do you explain uncertainty?
• Who should be involved?
• What about traffic from other AI assistants?
• Exercise
• Checklist
• From MentionX data: why AI mentions matter alongside lost clicks
• Worked example: Ledgerline forecasts the impact on its guides
• How do you check which queries show AI Overviews?
• How do you choose the time periods to compare?
• How do you separate AI impact from other changes?
• How do you set the scenario assumptions?
• How do you include conversions?
• What should the forecast lead to?
• How do you build the model in a spreadsheet?
• How do you check the forecast later?
• Common questions
• Related
• Sources

Checked on 31 October 2026. "Our view" marks MentionX opinion. Ledgerline is a fictional company used for illustration.

Why forecast at all?

Leaders plan budgets and headcount on traffic. Our view: a transparent forecast lets you shift effort early, for example from informational content towards pages that win citations and conversions, instead of explaining a drop after it happens.

What data do you need?

• More than a year of Search Console Performance data for the Web search type, which includes AI features (Google). The report defaults to three months, so change the date range (Google). Our view: export and store data monthly so you keep your own history.
• Landing page conversions from analytics.
• A manual sample of results pages showing which queries trigger AI Overviews.

See what Search Console shows for exports and filters.

Step 1: How do you group queries by exposure?

Use the manual sample to check how often AI Overviews appear in each group.

Step 2: How do you measure CTR change?

For each group, compare CTR year on year for queries where average position moved little. Our view: this isolates changes that aren't explained by ranking shifts. Note seasonality and other SERP changes.

Step 3: How do you build scenarios?

Our view: tie every assumption to something you've observed. If you haven't seen a change, say so and keep the low scenario near zero.

Step 4: How do you turn CTR into clicks?

Forecast clicks = expected impressions × scenario CTR, per group. Worked example (fictional): a high-exposure group with 1,000,000 impressions and CTR falling from 4.0% to 3.4% in the middle scenario gives 34,000 clicks instead of 40,000, a loss of 6,000.

Step 5: How do you translate to business impact?

Multiply lost clicks by the landing pages' conversion rate and value. Our view: informational pages often convert less, so the revenue impact may be smaller than the traffic impact. Google describes clicks from pages with AI Overviews as "higher quality" (Google); check whether your conversion rate on those pages has changed before assuming lost clicks were valuable.

How should you present it?

• A range, not a point.
• Assumptions per scenario.
• Which query groups drive the loss.
• What you'll do about it, and what that could recover.

Our view: include a line on what would change the forecast, such as new AI features or ranking changes.

What actions reduce the risk?

Our view:

• Make pages more useful than a summary: original data, tools, examples, first-hand experience. Google says unique, non-commodity content likely influences presence more than any other suggestion in its guide (Google).
• Aim to be a supporting link in AI Overviews; pages must be indexed and snippet-eligible (Google).
• Shift investment towards commercial and branded demand.
• Track AI visibility across other assistants, where buyers may also research.

Which pitfalls should you watch for?

• Applying a single external CTR figure to all queries.
• Ignoring ranking changes.
• Forecasting without a manual SERP sample.
• Presenting traffic loss as revenue loss one-for-one.

How often should you update it?

Our view: quarterly. Replace assumptions with observed data as it arrives, and keep the earlier forecasts so you can show how accurate they were. That builds trust in the next one.

How do you explain uncertainty?

Our view: say plainly which parts are measured and which are assumed. "We've measured CTR on question queries falling at stable position; we're assuming it continues" is honest and easy to follow. Avoid false precision, such as forecasts to the nearest click. Round to sensible figures and show the range on a simple chart.

Who should be involved?

Our view: SEO or content leads build the query groups, analytics checks the numbers, and finance reviews how traffic converts to revenue. A forecast built only by the SEO team is easy to dismiss; one checked by finance is easier to act on.

What about traffic from other AI assistants?

Our view: include a short note, even if you can't forecast it precisely. Some buyers move research to assistants such as ChatGPT, Perplexity or Copilot. Track referral sessions from those engines in analytics and mention rate in tracked prompts, and state whether they are growing alongside any search decline.

Exercise

Group your top 200 queries into high, medium and low exposure. Calculate year-on-year CTR change at stable position for each group and build low, middle and high click forecasts for the next 12 months.

Checklist

• Over a year of Web search data exported
• Queries grouped by exposure
• Manual SERP sample done
• CTR change measured at stable position
• Three scenarios with assumptions
• Converted to conversions and revenue

From MentionX data: why AI mentions matter alongside lost clicks

Method. Read-only aggregate queries over MentionX scans from 22 August to 29 October 2026 (UTC): 2,217 engine runs (2,154 completed) across 18 anonymised projects and six engines.

Clicks are only part of the picture. In the MentionX sample, Google's AI Mode cites sources in most answers, while other engines cite far less often:

Brands are also mentioned at very different rates depending on prompt type:

Our view: a traffic forecast should sit alongside a visibility measure. If clicks fall but your brand is mentioned and cited in answers, the business impact may be smaller than the traffic loss suggests.

Limits. A convenience sample of mostly B2B software projects, run through engine APIs, which can behave differently from the consumer apps. Perplexity, Copilot and DeepSeek figures come from a single day (26 October 2026), so treat them as a snapshot, not a trend.

Worked example: Ledgerline forecasts the impact on its guides

Ledgerline is a fictional accounts-payable tool. Figures are illustrative.

Ledgerline's head of growth wants to know how AI Overviews might affect traffic to its accounts-payable guides. She exports a year of Search Console data and groups queries into those that show AI Overviews often, sometimes or rarely, based on manual checks of a sample.

For each group, she compares click-through rates before and after AI Overviews appeared more often. She builds three scenarios with different rates of further decline. She converts the clicks into demo requests using the guides' historical conversion rate. The result is a range, not a single number, and it shows that most of the risk sits in definition queries, which convert least. All figures in the model are Ledgerline's own and hypothetical.

How do you check which queries show AI Overviews?

Search Console doesn't label AI Overview impressions separately, so you need another method. Check a sample of queries manually, signed out and from your main market, and record whether an AI Overview appears. Some rank-tracking tools also report AI features. Note that appearance varies by user, location and time, so treat any classification as approximate.

How do you choose the time periods to compare?

Compare periods of the same length and, where possible, the same season. Avoid periods with major site changes, migrations or Google updates unrelated to AI features. Record the dates you chose and why, so others can check the analysis later.

How do you separate AI impact from other changes?

Traffic changes for many reasons: rankings, seasonality, demand, site changes and competitors. Compare query groups that are exposed to AI Overviews with groups that aren't. If both fall, the cause may be something else. Check rankings for the same queries to rule out position changes.

How do you set the scenario assumptions?

Base the scenarios on your own observed CTR changes where you have them. Use the smallest change you've seen for the low scenario and the largest for the high one, with a middle case between. Write each assumption down. Avoid borrowing figures from industry studies without checking whether they fit your audience and queries.

How do you include conversions?

Use historical conversion rates for each page group, not a site-wide average. Informational pages usually convert less than commercial pages, so the same traffic loss can mean very different business impact. Where possible, include assisted conversions too.

What should the forecast lead to?

A forecast is useful only if it leads to decisions. Use it to prioritise which pages to improve for AI answers, where to invest in other channels and how to set targets for traffic and leads. Agree actions when you present it.

How do you build the model in a spreadsheet?

Use one row per query group, with columns for current clicks, current CTR, impressions, assumed CTR change in each scenario, forecast clicks, conversion rate and forecast conversions. Keep assumptions in a separate tab so they're easy to change. Add a notes column explaining each assumption. Our view: a simple, transparent model is easier to explain and update than a complex one, and it is usually just as accurate given the uncertainty.

How do you check the forecast later?

Each quarter, compare actual clicks with the scenarios. Note which scenario was closest and why. Adjust the assumptions and record the changes. Over time, this makes the forecast more reliable and builds trust with stakeholders.

Common questions

Can I forecast exactly how much traffic I'll lose? No; use scenarios and update them as data comes in.

Does Search Console show AI Overview clicks separately? Not as a separate filter; you need to group queries yourself.

How often should I update the forecast? Quarterly is a sensible default, or after major changes.

Related

• Search Console and AI Overviews: what the data shows
• Attribution: connecting AI visibility to pipeline

Sources

• Google, AI features and your website; Succeeding in AI search
• Search Console Help, Performance report

All accessed 30 Oct 2026.