The pattern is specific: positions in Google are the same or better, impressions are steady, but clicks are down and the phone is quieter. It is a real pattern, and it does have a likely cause. It also has four or five other likely causes that look identical in a dashboard, and the cost of guessing wrong is months of spend pointed at the wrong problem.
So diagnose first. The fix is cheap once the cause is known.
What is actually known about AI Overviews
Google launched AI Overviews in Search in May 2024, and measured click-through rates fall when an AI Overview appears above the results — that is how AdExchanger described the position in reporting published around January 2026.
On the size of the effect: research by Ahrefs, as reported by Search Engine Journal in approximately October 2025, found AI Overviews reduce clicks by 34.5% for the affected queries. Google has publicly taken the opposite position, that appearing in an AI Overview leads to higher click-through. Treat the 34.5% as a study finding on a sample of queries, not a law of nature.
A figure of "up to 47%" circulates widely in 2026 SEO guides. The underlying dataset is not disclosed. It should be treated as an agency claim rather than a measured benchmark, and anyone quoting it as a benchmark in a pitch is worth a follow-up question.
Outcomes vary enormously by business type. AdExchanger, citing Bloomberg, reported that travel site The Planet D lost roughly half its traffic in the months after the launch, while documented cases of businesses winning meaningful referral traffic from AI answers remain far rarer. That spread — catastrophic for some publishers, negligible for others — is the whole reason a generic answer is useless here.
Step one: check the enquiries, not the sessions
Before anything else, confirm there is a revenue problem.
Count actual enquiries — form submissions, calls, quote requests — for the last 90 days against the same 90 days last year. Not sessions. Not "leads" as reported by an analytics goal that may have broken.
For most local, trade and service businesses, the traffic most exposed to AI Overviews is informational: "how much does rewiring cost", "do I need a permit for a deck", "what is a franking credit". Those pages were rarely the ones that produced quotes. A 30% drop in sessions can sit alongside completely flat enquiry volume, because the queries that convert — "electrician [suburb]", the brand name, "emergency plumber near me" — behave differently to research queries.
If enquiries are flat and only sessions fell, there is a reporting problem and a content-strategy question, not a bleeding wound. That distinction is worth finding out in an afternoon.
Step two: rule out the boring causes
If enquiries genuinely fell, work through these before buying anything. The order below is not based on any measured data on which cause is most common — it is a rough sequence based on how quickly and cheaply each one can be ruled out. Work through all of them.
- The form is broken. Submit a test enquiry from a phone on mobile data, and from a desktop, and confirm the email arrives — including checking spam and any shared inbox rules. Do the same for the phone number as a click-to-call link. Broken forms after a theme or plugin update are common and produce exactly this pattern.
- Tracking changed. A new consent banner, a moved tag, a switch of analytics platform or a redirect chain will drop reported clicks and conversions without touching reality.
- Period mismatch. Comparing June against December, or a 28-day window against a 31-day one, is a known source of misattributed drops — Relevant Audience named seasonal comparison error specifically in a guide published around June 2026. Compare like for like: same months, same number of weekends, same public holidays.
- Paid or referral changes. A paused Google Ads campaign, a lapsed directory listing, a supplier who removed a link.
- Google Business Profile and Maps. For local businesses this can be the real story. Check whether the profile is still verified, whether categories or the service area changed, whether a new competitor appeared in the map pack, and what has happened to review volume and recency.
- A core update. Check the date the decline began against Google's published update history. A drop that starts on a single identifiable day usually is not a gradual SERP-feature effect.
Step three: the Search Console test
This is the part the checklists skip. In Google Search Console, compare a recent period against the matched period last year, and export queries with clicks, impressions and average position.
The AI Overview signature is: impressions flat or up, average position flat or better, clicks down — concentrated in question-shaped and informational queries, while brand and "service + suburb" queries hold steady.
If clicks fell across brand queries too, that is not a SERP feature. If position fell as well, that is a ranking problem, which is a different job.
Bear in mind other features absorb clicks the same way. Featured snippets, knowledge panels and People Also Ask were doing this before AI Overviews existed, as Relevant Audience noted in that June 2026 guide. Search the affected queries manually, on mobile, in a private window, and look at what is actually occupying the screen above the first blue link.
If it is AI Overviews: what helps
Be clear about the ceiling first. If an AI answer fully resolves the question, that click is gone and no amount of markup brings it back. The realistic goal is capturing a smaller volume of higher-intent clicks and being the business named in the answer. It is not restoring last year's session count.
A note on all of the below: none of the research above documents how AI Overviews choose which businesses to name. These are general local-search practices that are worth doing on their own merits, not a proven route into an AI answer.
- Be the same entity everywhere. Identical business name, address, phone, service area and category across the Business Profile, the website, directories and industry bodies. Inconsistent details are a plausible reason a business gets skipped, though the selection mechanics are not publicly documented.
- Reviews, continuously. Volume, recency and responses. Reviews are one of the few levers a local business fully controls, and they feed the Business Profile.
- Get cited by others. Local press, industry associations, supplier and manufacturer "find an installer" pages, chambers of commerce, genuine local directories. Being referenced by sources that are themselves cited is a reasonable bet, and useful regardless of what Google does.
- Answer the question on the page, in the first paragraph. Short, direct, specific to the location and the job. Then the detail underneath, for the reader who clicks through.
- Publish the things a summary cannot replace. Prices and price ranges, availability, service-area maps, before-and-after work, named staff, licence numbers.
- Structured data — with an honest caveat. Schema markup for organisation, local business, service, FAQ and reviews is sensible hygiene and helps machines read a page. The claims that it causes inclusion in AI answers are correlational. Have it done properly and cheaply; do not pay a premium for it as a guaranteed fix.
On the AEO and GEO retainers being pitched
A reasonable test: ask the agency to state, before signing, which queries lost clicks, what the matched-period comparison shows, and what the target is in enquiries rather than visibility. If the proposal quotes an undisclosed percentage and jumps straight to a markup checklist, it is selling a fix for a problem it has not measured.
For many businesses the honest answer is a smaller amount of work: fix the tracking, fix the form, tighten the Business Profile, build reviews, and accept that informational traffic is structurally lower than it was.
What to do this week
Submit a test enquiry. Pull enquiry counts year-on-year. Export Search Console queries for the matched period and check whether position held while clicks fell. Search the five queries that used to bring work and look at what sits above the results now. Only then decide whether there is a budget decision to make at all — and if the numbers point at a specific commercial choice, that is a conversation to have with someone who can see the full picture of the business.
Sources
- Search Engine Journal — Impact of AI Overviews: how publishers need to adapt (reporting Ahrefs study)
- AdExchanger — The AI search reckoning is dismantling open web traffic
- Relevant Audience — How to fix traffic loss from AI Overviews