Sitewide-drop recommendations find a sudden collapse of the whole site: clicks and impressions fall to half of their usual level or less on the same day, across most of the site’s biggest queries, and stay down. That pattern is what a technical fault looks like. A fall in demand is usually gradual, or hits some queries and not others.
To see it, open the site’s Recommendations report: press the lightbulb with a number in the top bar, or add the report to the site with Add report, From a template, Recommendations (see Dashboards & reports). The number on the lightbulb counts a fall of the whole site once. Groups are sorted by priority, so it usually appears at the top, under Check indexing, robots.txt and server errors: the whole site fell at once. The entry shows how many clicks and impressions the site has averaged a day since the fall, what is usual, and how many of the biggest queries fell together.
What to check first
The finding says that the whole site fell at once. It does not say why. Work through these in order, starting with the cheapest:
- Is the site up? Open it. Look for server errors, an expired certificate, DNS or hosting problems, or a content delivery network that stopped serving pages.
- robots.txt. Open
/robots.txt. A staging copy that went live withDisallow: /is a common cause after a launch or migration. - noindex. Check the page source and the response headers of a few pages for a
noindexrobots tag or anX-Robots-Tagheader applied to the whole site. - Redirects and canonicals. After a change of domain, address structure or HTTPS setup, check that pages still redirect to the right place and that canonical tags point to themselves.
- Sitemaps. Check that your sitemap still exists and lists the right addresses.
- Search Console itself. Look at the Page indexing report, the Crawl stats report, Manual actions and Security issues. This tool cannot see those: Google does not offer them through the API.
- Google’s side. Check Google’s Search Status Dashboard for an incident, and consider demand: a holiday or a major event can change search behavior for a few days.
Recovery can lag a fix, because Google has to crawl the site again before the numbers move.
What is compared
The rule works inside 28 consecutive completed days of Google Search Console web search data, the same interval the other query rules use. Both site totals and query data must have been fetched and marked final for every day, and the saved dates stay fixed when you change the report’s date picker.
Each of the last days is compared with the same weekday in the three weeks before the collapse, so a normal quiet weekend is not mistaken for a fall. “Usual” for a day is the median of those three days.
Preliminary days are excluded. If newer final data leaves a gap in the required coverage, the calculation waits for complete coverage. An interval that ended more than seven days before the calculation date is treated as stale, and a short warning appears when a listed finding belongs to a rule that needs newer data.
When a site qualifies
Every condition below must be met before a finding is saved.
| Condition | Current rule |
|---|---|
| Enough traffic | Before the last week, the site averaged at least 10 clicks and 100 impressions a day, and what is usual for the days of the fall is at least that much too. A quieter site cannot tell a fault from a quiet day, and this rule does not assess it. |
| It began suddenly | The collapse began within the last 7 completed days, and the day before it was ordinary: clicks and impressions each at least 80% of usual. |
| It has lasted | On every day from the first day of the collapse to the newest completed day, and for at least two days, clicks and impressions were at most 50% of usual and no higher than the lowest of the three comparable days. |
| It is broad | Of the site’s 30 biggest queries (by impressions before the collapse, brand queries included), at least 70% lost at least half of their daily impressions over the same days. At least 10 such queries are needed. |
A single bad day is not enough, because a data hiccup looks the same. The rule therefore reports a collapse at the earliest on its second day, and Search Console data runs a few days behind, so a finding appears a few days after the fall began.
Reading the estimate
The clicks at risk are calculated as:
(usual clicks per day − actual clicks per day) × 28
Both are averages over the days of the collapse. For example, a site that usually earns 320 clicks a day and now earns 25 has about 8,000 clicks at risk over 28 days if it stays down.
The estimate answers “what is at stake if the site keeps losing clicks at this rate”. It is not a forecast and does not predict how fast the site recovers.
Each finding’s priority is clicks at risk × evidence strength. Evidence strength is a score from 0.5 to 1 that grows with how deep the fall is, how many of the biggest queries fell together and how much traffic the site normally has. It is not a probability that the cause is technical or that traffic returns. A collapse of the whole site usually outranks the other findings.
Why this recommendation? shows clicks and impressions for each day of the collapse next to what is usual for that weekday. It also names the day before the fall by its date and says that it was ordinary.
Which findings appear
All active findings appear in the list. The report has no dismissal or snooze controls. When a later completed calculation no longer shows the collapse, because traffic came back or because the conditions no longer hold, the finding becomes resolved and leaves the list. If it qualifies again, it returns.
Saved results and limitations
Calculations run after sync. The findings, rule version, threshold values and numerical evidence are saved. Opening the report reads the saved results and does not run the rule again or change Search Console data. If newer data is incomplete, the earlier result stays visible and the report notes that some suggestions use older data.
- It covers the first week. After seven days the fall is no longer sudden in the data, because the days before it are also down, and the finding resolves even if the site has not recovered. The other recommendations keep showing what the fall cost.
- Days with no rows count as zero. Search Console returns no rows for a day with no impressions, so a site that disappears from results entirely looks like a fall to zero.
- It sees the numbers, not the cause. A data delay, a holiday, an outage on Google’s side and a real technical fault can look the same. Check before you change content.
- Anonymized queries are missing. Google omits them from query data, so the breadth check uses the queries it receives. See Data & limits.
- Web search only. Image, video, news and Discover results are not included.
If only some queries or pages fell, see position-drop recommendations and page click-drop recommendations.