Your rankings did not move, but a Google AI Overviews traffic drop may explain why your traffic did.
That gap has been landing on New Zealand and Australian websites since AI Overviews rolled out to more than a hundred countries. Same keywords. Same positions. Fewer people arriving. Search Console shows impressions holding steady while the clicks line slides down and never comes back.
Most owners I talk to assume they have been penalised. They have not. Google put an answer at the top of the page, and for a large share of searches, that answer is now enough.
This post covers what actually changed, how to read your own numbers instead of guessing, which queries are worth fighting for, how to become the source the AI quotes, and what to do about the visitors you still get.
Key Takeaways
- A Google AI Overviews traffic drop usually shows up as flat impressions, falling clicks and unchanged rankings in Search Console.
- Pew Research found people clicked a search result on 8% of visits where an AI summary appeared, against 15% of visits without one.
- Informational and definitional queries bleed first. Pricing, comparison, local and booking queries hold up far better.
- Getting quoted in an AI Overview needs a direct answer in the first 40 words of a section, not a 300-word warm-up.
- AI assistants cite what other sites say about you, so reviews, directories and third-party mentions now feed your search visibility.
- Schema markup and clean FAQ blocks hand AI systems a machine-readable version of your answer, which raises the odds of citation.
- Lower traffic makes conversion the fix. A missed call or a four-hour reply wastes visitors you can no longer afford to lose.
- One finance client recovered $49,000 from 319 dormant contacts, revenue that never depended on search traffic at all.
What a Google AI Overviews Traffic Drop Actually Looks Like
The reason this catches people out is that it does not look like a normal traffic problem. There is no manual action, no algorithm update announcement, no ranking collapse. Everything on the diagnostic checklist comes back clean.
Here is the signature.
Impressions flat, clicks down, position unchanged
Open Google Search Console, go to Performance, set the date range to the last 16 months, and turn on all four metrics at once: clicks, impressions, average CTR, average position.
If the drop is caused by AI Overviews, you will see impressions roughly level or slightly up, average position roughly level, and clicks and CTR falling together. Your pages are still being shown. People are just not clicking through because they got what they needed before they reached your listing.
Compare that to a genuine ranking problem, where impressions and position both fall at the same time. Or a seasonal dip, where impressions and clicks fall together, but CTR holds. The pattern tells you which problem you have. Guessing does not.
The numbers behind it
The independent research is consistent enough to plan around. Pew Research Centre tracked real browsing behaviour and found users clicked a traditional search result on 8% of visits where an AI summary was present, compared with 15% where it was not. Roughly half the clicks, gone. They also found that only about 1% of users clicked a link inside the AI summary itself.
Independent SEO studies have landed in the same territory, reporting click-through rate falls in the 30% to 40% range for top-ranking pages once an AI Overview appears above them.
So a 30% to 50% drop in organic clicks on affected queries is not a sign that something is broken. It is the current maths of the page.
It is not evenly spread
This is the part that matters for planning. The drop is not spread evenly across your site. A handful of pages will be down 60% while others have barely moved.
Sort your Search Console pages by click loss year on year, and you will almost always find the same split: your explainer content is haemorrhaging, and your money pages are close to flat. That split is the whole strategy, and I will come back to it.
Why Your Rankings Held While Your Clicks Fell
Nothing happened to your site. Something happened to the page your site sits on.
The result you ranked for moved down the screen
Position one used to mean the first thing a person saw. On a query that triggers an AI Overview, position one now sits below a generated answer that can run several hundred words, complete with a handful of cited links off to the side.
You did not lose the ranking. You lost the screen real estate above it. On a phone, that can mean two or three thumb-scrolls before your listing appears, and most people never scroll.
Zero-click search was already eating your lunch
AI Overviews did not invent this. Featured snippets, local packs, knowledge panels, People Also Ask boxes, and the answer-in-the-SERP behaviour have been chipping away at click-through rates for the better part of a decade. Zero-click searches were already the majority of Google sessions before generative answers arrived.
What changed is the scale and the confidence. A featured snippet lifted one sentence from one page. An AI Overview synthesises several sources into a complete answer that reads like it settled the question. For a query like “what is answer engine optimisation” or “how long does a bathroom renovation take”, it usually has.
Google’s position, and what is actually true
Google’s line is that AI Overviews send higher quality clicks, and that overall click volume to the web is stable. Publishers with traffic dashboards in front of them disagree loudly.
Both things can be partly true. The clicks that survive an AI Overview tend to be from people who wanted more than a summary, which means they are further along and more likely to act. That is genuinely better traffic. It is also less of it.
Planning around “the clicks are better quality” is not a strategy. Planning around “I get fewer visitors, and each one has to count for more” is.
Which Traffic to Fight For and Which to Let Go
The instinct after a traffic drop is to write more blog posts. That is usually the wrong move, because the posts most people write are exactly the ones AI Overviews absorb best.
Split your pages by intent, not by traffic
Export your Search Console query data and sort every query into two buckets.
Bucket one: the ones AI answers well. Definitions, “what is”, “how does X work”, general how-to, statistics, comparisons of concepts. These are informational, answerable in a paragraph, and have no commercial edge. An AI Overview handles them completely. You will not win these back with a longer article, and you should stop spending your best hours on them.
Bucket two: the ones AI cannot finish. Anything where the searcher needs to make a decision, spend money, check availability, or trust a specific person. Pricing, “near me”, “best X in Auckland”, reviews, comparisons of named providers, booking, and anything with a local or personal dimension. A generated summary can tee these up. It cannot close them.
The second bucket is where your remaining budget goes.
Sharpen the bottom of the funnel
Most small business sites are thin exactly where they should be thick. Ten blog posts explaining what a service is, and one weak service page that does not name a location, list a price range, answer an objection or show a single review.
Fix that order. A service page that names the suburb, states what a job typically costs, shows real photos of real work, answers the five questions every prospect asks and carries genuine reviews is very hard for an AI summary to replace. It is also the page that converts.
Stop measuring the wrong thing
If you have been reporting on sessions, that metric is going to keep telling you a depressing and mostly useless story.
Move to enquiries per month, cost per enquiry, and conversion rate from enquiry to job. If sessions fall 30% and enquiries stay flat, nothing bad happened to your business. If sessions fall 30% and enquiries fall 30%, you have a real problem and now you can see it clearly. That distinction is impossible to make from a traffic chart alone, which is one of the reasons I push every business I work with to get their numbers into one place rather than six dashboards. More on that in AI adoption for SMEs.
How to Become the Source AI Overviews Quotes
Losing clicks to an AI Overview is bad. Being invisible inside it is worse. Citation in the summary still drives some traffic, and more importantly, it puts your name in front of a person at the exact moment they are forming an opinion.
This is what people mean by answer engine optimisation, and it is less exotic than it sounds. It is mostly good writing plus structure.
Answer the question in the first 40 words
Generative systems lift passages, not pages. If the answer to the heading is buried in paragraph four after a warm-up about how the industry has changed in recent years, the model has nothing clean to take.
Every H2 and H3 on your site should be a question or a clear claim, and the first two sentences underneath it should answer that question completely. Then expand. A reader who wants the short version gets it. A model that wants a quotable passage gets one too.
Test it yourself: read the first 40 words under each heading in isolation. If they do not stand up as a complete answer, rewrite them.
Make each section liftable
Short paragraphs. One idea per paragraph. Specific numbers instead of vague ranges. Named entities instead of pronouns. If a section says “it usually takes about a week”, change it to “a standard bathroom renovation in Auckland usually takes 10 to 14 working days”.
Specificity does two things at once. It gives the AI a fact worth citing rather than filler worth ignoring, and it gives the human reader a reason to believe you.
Add the schema
Schema markup is the machine-readable version of your page. FAQPage schema on your FAQ sections, LocalBusiness schema on your contact and service pages, Article schema on your posts, Review schema where you have genuine reviews.
None of this is a ranking trick. It is a translation layer. You are removing ambiguity about what your page says, who wrote it and what business it belongs to, which makes you a safer thing to quote.
Get mentioned where the models already look
Here is the shift most people miss. AI answers do not just read your website. They read what the rest of the internet says about you.
That means Google Business Profile completeness, review volume and recency, industry directories, local news mentions, podcast appearances, LinkedIn presence and any third-party listicle that ranks providers in your category. If an assistant is asked “who is the best commercial electrician in Hamilton”, it is synthesising from those sources, not from your homepage copy.
Chasing that surface deliberately is a different job from traditional SEO, and I have covered the mechanics of it in how to get recommended by ChatGPT and answer engine optimisation for small business.
Track the assistants, not just Google
Google is one channel now. ChatGPT, Claude, Perplexity, Copilot and Gemini all send referral traffic, and for some categories it is already meaningful.
Set up a monthly check. Ask each assistant the five questions a good prospect would ask before hiring you. “Best [your service] in [your city].” “Who should I use for [problem].” “Is [your company] any good?” Write down whether you appear, who appears instead, and what the assistant says about you. Then check your analytics referral report for traffic from those domains.
Fifteen minutes a month. It is the only visibility report that matters for this channel, and almost nobody is doing it.
Fewer Visitors Means Every Enquiry Has to Count
Everything above takes months to compound. Here is the part you can fix this fortnight, and it is worth more than the rankings.
The maths changed at the top, so fix the bottom
If your website used to bring 1,000 visitors and 20 enquiries, and it now brings 700 visitors and 14 enquiries, you have two ways back to 20. Win back 300 visitors, which is slow, expensive and partly out of your control. Or convert better, which is fast, cheap and entirely in your control.
Most founder-led businesses have enormous slack here and cannot see it, because nobody measures the leaks.
Answer the phone
The single biggest leak I find is the one nobody records. Calls that ring out. Calls that go to voicemail at 4:50 pm. Calls that come in while the one person who answers phones is with a customer.
A dental practice I worked with was losing 47% of its inbound calls despite having two receptionists. Not because the team was slow, but because the calls came in clusters and the phones do not queue politely. After an AI receptionist picked up the overflow, missed calls went to zero and booked appointments were up 44%.
That is not a marketing gain. That is the same demand, captured instead of lost. Worth reading alongside after-hours call answering if your phone rings outside business hours.
Reply in minutes, not hours
Roughly 78% of deals go to whoever responds first. Most small businesses take four hours or more to make first contact with a web enquiry, usually because the person who handles it is on a job site or in a meeting.
An enquiry that gets a real response in 90 seconds converts at a completely different rate to one that gets a response tomorrow morning. And the fix is not “try harder”. It is a system that contacts every new enquiry immediately, qualifies it, and books the call, whether or not you are at your desk.
When your traffic drops 30%, closing the response gap can recover more revenue than the traffic was worth in the first place.
Work the database you already paid for
Every established business has a list of people who enquired, got a quote, went quiet and were never followed up properly. That list cost you money to build. It is sitting there earning nothing.
A finance broker I worked with had 319 dormant contacts his team had written off completely. A multi-touch reactivation across SMS and email, written to sound like a person rather than a campaign, recovered $49,000. No ad spend. No new traffic. No dependency on Google whatsoever.
That is the point worth sitting with. The revenue in that database was never exposed to an AI Overviews traffic drop, because it never came through search in the first place.
The bigger pattern
A drop like this exposes something that was already true. If your business depends on one channel behaving the way it did last year, you are one product decision at a large company away from a bad quarter.
The businesses that shrug this off are the ones where the phone gets answered, enquiries get responded to fast, the database gets worked systematically, and the owner can see all of it in one place instead of piecing it together from six dashboards. That is an AI brain that knows the business plus an AI workforce that does the repetitive work, and it is a far more durable asset than a ranking.
Conclusion
A Google AI Overviews traffic drop is not a penalty, and it is not reversible by writing more of what you were already writing. Google put an answer above your listing, and for informational queries, that answer is now enough for most people.
The response has three parts. Read your own Search Console data properly so you know which queries actually moved. Rebuild your content so it is quotable, specific and structured for machines as well as readers, and go and earn mentions on the sources those machines already trust. Then close the gaps at the bottom of the funnel, because fewer visitors means every phone call, every form fill and every dormant contact carries more weight than it did a year ago.
The traffic is not coming back in its old shape. The revenue can.
Book a Discovery Call
If your traffic is down and you are not sure whether your enquiries have followed it, that is worth 30 minutes.
Book a Discovery Call and I will look at where your enquiries currently come from, where they leak, and what would move the number fastest. No pitch deck, no obligation, and you will leave with at least one thing you can fix yourself that week.
Frequently Asked Questions
Why has my website traffic dropped even though my rankings are the same?
Because the page changed, not your position. When Google shows an AI Overview above the organic results, your listing gets pushed further down the screen, and fewer people scroll to it. In Search Console, this appears as flat impressions and average position, with clicks and click-through rate falling together. That pattern points to AI Overviews rather than a ranking or penalty problem.
Do AI Overviews actually reduce clicks to websites?
Yes, measurably. Pew Research Centre found people clicked a traditional search result on 8% of visits where an AI summary appeared, compared with 15% where none did. Independent SEO studies have reported click-through rate falls of roughly 30% to 40% for top-ranking pages when an AI Overview is present. The effect is strongest on informational queries and much weaker on commercial ones.
Can I stop Google using my content in AI Overviews?
Only by using snippet controls like nosnippet or max-snippet:0, and those also remove you from featured snippets and normal search snippets, which usually costs more traffic than it saves. The Google-Extended control affects Gemini training rather than AI Overviews in Search. For almost every small business, the better play is to be cited well rather than to opt out.
Can you rank inside a Google AI Overview?
There is no ranking to buy or bid on, but you can raise the odds of being cited. Answer the question directly in the first 40 words of a section, use clear question-style headings, be specific with numbers and named entities, add FAQPage and LocalBusiness schema, and build mentions on the third-party sites and directories these systems already read when forming an answer.
Is SEO dead because of AI Overviews?
No, but the target moved. Optimising for informational keywords that an AI can fully answer is now poor value. Optimising for commercial, local and decision-stage queries still works, because a summary cannot check your availability, quote your job or earn your prospect’s trust. The bigger change is that visibility now also depends on what other websites say about you.
How do I check if AI Overviews are affecting my site?
Open Google Search Console, set the date range to 16 months and display clicks, impressions, click-through rate and average position together. If impressions and position are steady while clicks and CTR fall, AI Overviews are the likely cause. Then sort by page and by query to see exactly which content lost ground, and search those queries yourself to confirm an AI Overview appears.
Should I still write blog posts if AI answers the question anyway?
Write fewer and make them better. Generic explainer posts that an AI Overview fully answers are no longer worth the hours. Posts with original data, real client examples, local specifics, strong opinions and genuine expertise still earn links, citations and trust. Then put your remaining effort into service, pricing and location pages, which convert and which AI cannot replace.
About Octavius
Titus Mulquiney is the founder of Octavius AI, where he builds AI brains and AI workforces for founder-led businesses stuck running everything out of their own head. Twenty years in marketing, ex-Sony product manager, ex-GM Zeal NZ. Based in Auckland, working with operators across NZ, Australia, and the US. Connect on LinkedIn.