Are you comparing the same thing?
Use complete periods of equal length, with similar days of the week. Check the same time last year if the business is seasonal. Note holidays, opening hours, stock, service availability and promotions.
Keep the actual numbers beside the percentage. A fall from 20 Google clicks to 14 is 30%, but it is six clicks. Six clicks can matter; they do not establish six lost customers. A small count can also move sharply without a lasting trend.
Distinguish all website visits from visits attributed to Google. Your analytics tool may include email, social, paid advertising, other websites and direct or unknown sources. Search Console measures Google Search activity, so it answers a narrower question.
Could the tracking have changed?
Check whether you recently changed the website, analytics tag or cookie choices. If analytics falls while Search Console clicks remain broadly stable, investigate collection and reporting before assuming people stopped finding you.
Ask whoever maintains the website to trace a normal permitted visit and meaningful enquiry or order through the intended reports. Respect customers' consent choices. Do not repair a measurement gap by labelling every unclassified visit as AI or a customer.
Check your actual inbox, booking system and payments too. A dashboard event called “contact” may record opening the contact page rather than sending a message. Decide using the result you really need.
Can a tracking fault hide a real search problem?
Consider this invented comparison of two complete four-week periods for a service business. Search Console shows 100 Google clicks in each period. Its page breakdown changes: the main service page falls from 40 clicks to ten, while the guides rise from 60 to 90. Website analytics records Google-labelled sessions falling from 80 to 50 after a redesign. The actual inbox contains four suitable enquiries before and one after.
The stable Google total hides thirty lost service-page clicks, offset by thirty additional guide clicks. The guide gains may be useful, but they cannot establish that the lost service opportunities have been replaced. The enquiry change deserves attention without supplying a reliable conversion rate from these small, differently defined records. Inspect the affected pages and their queries in the same country and device; total Google clicks alone cannot close the diagnosis.
Now suppose two checks find separate faults. The important service page carries an unintended noindex instruction, which tells search services not to include it. On the new template, analytics also fails to load during a clearly labelled test for which analytics consent was given. These are constructed findings, not faults we have observed on a customer's website. Finding the tracking error does not make the service-page finding disappear, and the numbers alone did not establish either fault.
Give the provider two bounded repair jobs. Confirm the intended public service page's indexing eligibility after correcting the accidental exclusion, then allow for Google's recrawl and reassessment. Separately repeat the permitted measurement test, confirm its intended record once and exclude that test from customer totals. Review the service-page search evidence and suitable enquiries over the next comparable period. Do not declare the whole decline a tracking illusion after fixing only the tag, or buy a sitewide AI-traffic recovery package from this evidence. Google's page-level traffic diagnosis
If Google visits fell, what should you check?
In Search Console, compare the periods and look at Pages and Queries. Identify the pages that lost clicks and the searches associated with them. Check your main country and mobile traffic, rather than mixing markets you do not serve.
An impression records a search appearance under the report's rules; a click records selection of a result. Average position summarises where results appeared, not a permanent rank for the business. Use them together to investigate the change.
Use URL Inspection on an important affected page. Is the correct public address accessible? Has a redesign accidentally excluded it, redirected it elsewhere or changed which address is treated as the main version? A website provider can help investigate these findings. Open the Security issues and Manual actions reports as part of the check; an empty report does not rule out a ranking change.
If impressions fell, investigate demand and visibility. If relevant positions weakened, compare the page with the current competing answers. If impressions are similar but clicks fell, inspect the displayed title, other results and the search features present. These are useful clues, not automatic diagnoses. Google's traffic-drop guidance
Try arriving through an actual Google result on your phone, as well as opening the page directly. Check the address you reach, the offer and the next step. A page may load correctly yet leave a prospective customer unable to establish the terms or send an enquiry.
For a local business, inspect available Business Profile performance on Search and Maps, including website-link clicks, and check that hours and service details remain correct. Those clicks are not confirmed website sessions. Google’s metric definitions
Google Trends can provide context for changing interest in a broader term; sparse local queries may not produce useful data. If you need a specialist, bring the dates, affected URLs, channel and search comparisons, release history and enquiry record.
Is AI taking the clicks?
Sometimes an answer on the results page can reduce the need to visit a website. That possibility does not identify the cause of your own decline. Look at a few of the actual searches that lost clicks and record the date, market and results you see. Keep this as a sample.
Consider what customers still need from you after a short summary: the right service for their circumstances, a genuine example, current availability, a quote or a booking. Improve that part of the page. There is little value in hiding basic information and making people work harder to discover whether you can help.
Check whether suitable enquiries or contribution have fallen alongside traffic. If enquiries hold steady, there may still be a future discovery risk, but a panic purchase of more visits is not justified by the count alone.
What should you do next?
Choose one evidenced job: repair an excluded page, correct an inaccurate offer, investigate a lost query group or test a useful existing channel. Record what you expect to improve and when you will review it. Repeatedly changing the whole site makes learning harder.
Use email, partnerships or another appropriate channel to reduce reliance on one source when you have a reason to believe your customers use it. Diversification does not fix missing tracking or a broken form. Start small and compare suitable enquiries, retained customers and the effort required, rather than matching an old visit total at any cost.
Record the action the diagnosis supports and the attractive-looking fixes it allows you to defer in a short weekly business record. Review the longer trend before committing more money.