Your positions look the same. Impressions look healthy. But the enquiry form has gone quiet, and the jobs booked this month don't match what the reporting says is happening.
The usual suspect right now is AI-generated answer summaries on the results page — the panel that answers the searcher's question above the links, so the click never happens. That is a real mechanism, and it may well be part of what you are seeing. It is not the only thing that produces this pattern, and it is not the first thing to check.
Before changing anything on the site, work out which of three things is actually happening. They look identical in a monthly summary and they need completely different responses:
- Fewer clicks per impression. People are seeing the listing and not clicking — because an AI answer summary, a featured snippet, ads or a competitor's richer listing has already dealt with them.
- A tracking or form failure. Enquiries are being sent and not arriving, or arriving and not being counted.
- Seasonality or demand change. Fewer people are searching for the thing at all, or they are searching with different words.
Most advice skips straight to fix number one. That is expensive if the real problem is a broken form.
Step one: prove the form still works (do this today)
This takes twenty minutes and rules out the most embarrassing cause.
- Submit a test enquiry from a phone on mobile data, not office wi-fi. Use a real-looking name and a personal email address.
- Check whether it lands in the inbox, the junk folder, the CRM, and any autoresponder — all four, separately.
- Ring the number listed on the site and on the maps listing. Confirm it rings the right handset.
- Check for missed or unanswered calls in the phone system. A drop in "enquiries" is sometimes a drop in calls answered.
- Look at the date any of this last changed: a new theme, a plugin update, a cookie or consent banner, a change of email host, a new spam filter. Line that date up against the date enquiries fell.
If the site records conversions, test whether the conversion is still firing, not just whether the email arrives. Consent banners and tag changes routinely stop measurement without stopping the form. That produces a reported drop with no real drop — worth knowing before anyone panics.
Step two: read impressions against clicks, not clicks alone
For Google organic, Search Console is the practical place to separate "people saw us" from "people came to us". Bing Webmaster Tools does the same job for Bing, and ad platforms report their own impressions and clicks separately. If a meaningful share of the traffic comes from somewhere other than Google organic, check that source in its own tool rather than assuming one dashboard covers everything.
Pull the last 16 months, compare the same months year on year rather than month against previous month, and look at four numbers together: impressions, clicks, average position, and click-through rate.
The pattern tells the story:
- Impressions flat or up, position flat, clicks down, CTR down. Visibility is intact; the click is not happening. Something on the results page is satisfying the searcher before they reach the site. Note the limit here: Search Console reports impressions in aggregate and does not separate appearances alongside an AI-generated answer from ordinary listings, so this pattern is consistent with AI answers absorbing the click but is not proof of it. More ads above the organic results, a competitor with a richer snippet, or a changed title and description produce exactly the same shape.
- Impressions down, position flat, CTR flat. Fewer searches, or fewer searches where the site is eligible. That is demand or seasonality, not a penalty.
- Clicks flat, sessions flat, enquiries down. The problem is on the site or in the tracking, not in search.
- Position down along with everything else. That is a ranking issue and deserves the usual technical and content investigation.
Then go down a level. Sort queries by lost clicks. Informational queries — "how much does X cost", "what is Y", "do I need Z" — behave differently from queries with intent to hire, like a service plus a suburb. If the losses are concentrated in the first group and the second group is stable, the channel has changed shape rather than collapsed. If the hiring queries are also down, look harder at the listing, the maps results and paid competitors sitting above the organic results.
It is also worth running the losing queries yourself, on a phone, logged out. What you see on the results page will tell you more about the cause than the dashboard can.
Do the same page by page. A blog post losing clicks and a service page losing clicks are not the same problem.
Step three: rule out seasonality honestly
Compare the same period last year and the year before, in the same business. Then compare against something that does not depend on search at all — inbound calls, referral volume, quote requests from repeat clients. If those are soft too, the market is soft, and no amount of site work will fix it. Trades, professional services and anything tied to school terms or the end of the financial year all have shapes that repeat. Whether that applies to a particular business depends on its own history, which is why the year-on-year comparison matters more than any general rule.
What to do when the click really has gone
If the diagnosis is fewer clicks per impression — whether from AI-generated answer summaries, snippets, ads or all three — the honest position is this: a share of the traffic that used to arrive for general questions is not coming back. Arguing with that is not a strategy. Two responses are available, and a local service business should run both.
Shift weight to channels that do not need a click.
- The business listing on maps and search does the work of a landing page: hours, service area, photos, reviews, a phone number people can tap. Keep it current and keep reviews arriving steadily rather than in bursts.
- Email. A list of past customers and enquirers is owned, not rented. Even a short monthly note keeps the business in mind when the need arrives.
- Referral relationships with adjacent trades and professionals. These convert without ever touching a results page.
- Repeat and reactivation. A call to last year's customers usually beats a month of content.
Make the business the answer rather than competing with the answer.
General explanations are exactly the material an AI answer summary can produce without sending anyone anywhere. What it cannot produce is anything only this business holds:
- Real price ranges, or at least the variables that move the price, stated plainly on the service page.
- Specific service area, response times, availability and what the business does not do.
- Photos of actual jobs, with the problem, the constraint and the outcome.
- Named local detail — council requirements, common building stock, site conditions — that a generic summary cannot produce.
- Clear, structured facts on the page: business name, address, phone, hours, services, FAQs answered directly and briefly.
Then shorten the path for the people who do click. High-intent visitors arriving on a service page should be able to enquire in one screen: a phone number that dials, a short form, a booking option if the business takes bookings.
Change what gets measured
Sessions are becoming a weaker proxy for interest. Move the reporting to outcomes that survive the change:
- Enquiries by source, with a "how did you hear about us" field or question.
- Calls, including missed ones.
- Quotes issued and jobs won.
- Branded search impressions — people looking for the business by name is a reasonable indicator that awareness is holding even while clicks fall.
This week
Test the form and the phone. Pull 16 months of Search Console — plus Bing Webmaster Tools and any ad platform that matters to you — and compare year on year. Split queries into informational and hiring intent. Run the losing queries yourself and look at what the results page now shows. Check whether non-search channels moved at the same time. Only then decide whether the problem is measurement, demand, or the results page — and brief anyone doing marketing work on the answer rather than on the symptom.
If the numbers point somewhere ambiguous, a second set of eyes on the raw data is worth more than a strategy document written before the diagnosis.
Sources
No external data sources were supplied for this article, and no traffic-decline statistics are quoted. All figures referred to here are ones the reader should pull from their own Search Console, Bing Webmaster Tools, analytics, ad platforms, phone system and CRM records for their own business.