The pattern most owners are seeing looks like this: impressions in Google Search Console are flat or up, clicks are down, and the average position hasn't moved much. Nothing obvious broke. The site still ranks. Fewer people arrive.
Before spending a dollar on a new retainer, it is worth separating three very different problems that produce the same-looking graph: a genuine ranking loss, a seasonal or demand change, and clicks being absorbed by search features that answer the question on the results page. They need different responses, and only one of them is about "optimising for AI".
What is actually known, and what is not
Google publishes guidance for site owners on how content is surfaced in AI Overviews and AI Mode. According to Google Search Central's documentation on AI features, the standard Search technical requirements and indexing rules apply, and there is no separate submission process for these features. Check the "last updated" date shown at the top of that page when reading it — the documentation changes.
AI Overviews have been reported as rolling out to Australian users from 2024, with expansion across more query types since. Availability and which query types are affected vary by market and over time, so confirm the current position from Google's own release notes rather than from agency commentary or a US case study.
What follows from that: there is no submission form, no verification badge, no "AI ranking" to buy. Anyone selling one is selling a guess.
Step one: prove it in Search Console before you buy anything
Google Search Console Help states that its performance reports include impressions and clicks from surfaces where AI features appear. That makes the impressions-versus-clicks comparison the practical first test.
Do this in about an hour:
- Open Search results in Search Console. Set the date range to the last 3 months and compare to the previous 3 months, or to the same period last year if the business is seasonal.
- Turn on all four metrics: clicks, impressions, average CTR, average position.
- Read the shape:
- Impressions flat or up, clicks down, position roughly stable — the pages are still being shown, and fewer people are clicking. Consistent with results-page answers absorbing the click, though not proof on its own.
- Impressions down and position down — an ordinary ranking loss. Treat it as a ranking problem: competitors, technical issues, a content or link change, or an algorithm update.
- Impressions down, position stable — demand has fallen. Fewer people are searching. That is a market or seasonality issue, not a website issue.
Step two: split it by query type and by page
The aggregate number hides the answer. Split it.
- By query. Sort queries by click loss. Informational queries — "how much does X cost", "what is Y", "do I need Z" — are the ones most exposed to being answered on the results page. Transactional and brand queries usually behave differently. If the loss is concentrated in how-to and definition queries while enquiry-intent and branded queries hold, that is a meaningful pattern.
- By page. Filter to individual landing pages. If blog and guide pages lost clicks while service pages, pricing pages and the contact page held steady, the traffic that disappeared was the traffic least likely to become an enquiry. Painful for a traffic report, less painful for the business.
- By device and country. Rule out the boring explanations. A drop confined to one country or one device usually has a mundane cause.
Then compare against the number that actually matters: enquiries, calls, bookings and quotes over the same period. To illustrate the point with made-up numbers: a 30% traffic drop with a flat enquiry count is a different situation to a 30% traffic drop with a 30% enquiry drop, and the two should lead to different budget decisions. Which of those you are in is an empirical question about your own business — check your own enquiry records rather than assuming either outcome.
Rule out the ordinary causes
Before concluding anything about search features, check the unglamorous list:
- A site migration, redesign or URL change in the affected period.
- Tracking changes — a consent banner, a broken analytics tag, a bot-filtering change.
- Pages accidentally set to `noindex`, or blocked in robots.txt.
- A competitor that started bidding on your brand terms.
- Google Business Profile changes: suspensions, category edits, hours, or a drop in review volume. For a local service business this often moves the phone more than the website does.
What improves the odds of being cited
No one can guarantee a business appears in a generated answer, and no one can "rank number one in ChatGPT". There is no ranked list to be first in. What can be influenced is the odds, and per Google's own AI-features guidance, those odds run through ordinary indexable, crawlable content — not a separate channel.
Practical work, in rough order of return for a small business:
- Answer the specific question on the page, near the top. A clear two-to-three sentence answer to the exact question, followed by the detail. Buried answers are hard to quote.
- Publish the facts only you have. Prices or price ranges, service areas, lead times, what is and isn't included, qualifications, licences, warranty terms. Generic advice content is replaceable. Specifics about your business are not.
- Fix the technical basics. If Googlebot cannot crawl and index a page, nothing downstream matters.
- Get the business data consistent everywhere. Name, address, phone, hours, service area and categories — on the website, on Google Business Profile, and on the directories that matter in your industry. Contradictory data is a reason to be skipped.
- Keep reviews current. For local service businesses this generally does more than blog volume.
- Structured data on the pages where it applies, following Google's published requirements.
The controls, and the trade-off
Google Search Central documents that snippet use can be controlled with existing robots meta directives — `nosnippet`, `max-snippet` and `data-nosnippet` — and that Google-Extended is separate from ordinary Googlebot crawling for Search. Verify the current behaviour and the page's last-updated date before acting.
Be careful here. Restricting snippets can reduce how content is displayed in ordinary results too. Blocking is a defensible choice for some publishers whose product is the text. For most service businesses that want to be found, it works against the goal. Whether it suits a particular business depends on how that business makes money, so treat it as a decision to think through rather than a default.
What to do this week
Run the Search Console comparison and write down which of the three shapes it matches. Split it by query and by page. Put the enquiry numbers next to the traffic numbers. Only then decide whether there is a problem worth spending on — and if there is, whether it sits with the website, the Google Business Profile, or the offer itself.
If a proposal arrives promising a position in generated answers, ask what the deliverable actually is. If the answer is indexable content, clean technical foundations and accurate business data, that is legitimate work under a new name. If the answer is a ranking guarantee, it cannot be honoured.
Sources
- Google Search Central — AI features and your website
- Google Search Central — Overview of Google crawlers and user-triggered fetchers
- Google Search Console Help — Performance report
- Google — The Keyword, Search product announcements