Diagnosing a traffic drop

Traffic fell. The instinct is to look for the interesting explanation — an algorithm update, a competitor, a penalty. The discipline is to rule out the boring ones first, because the boring ones are far more common and much cheaper to check.

Work down this list in order. Stop when something explains the size and shape of the drop.

1. Is the drop real?

Check the date range against the same period last year and the preceding weeks. Seasonal businesses drop every August. Also check that the last few days are not just Search Console's incomplete recent data.

2. Did measurement break?

An analytics tag removed in a deploy, a consent banner change, a domain migration, a broken container. If clicks in Search Console are flat while analytics sessions fell, the tracking broke — not the traffic. This is the single most common cause of a scary chart.

3. Did the site break?

Check indexing: is the affected section still indexed? A noindex shipped by accident, a robots.txt change, a canonical pointing somewhere wrong, a template that stopped rendering content, a server returning 5xx to Googlebot. Search Console's Page indexing report is the fastest read here.

4. Was there an incident?

Google itself has bad days. On 28 August 2026 Search had an indexing and serving incident where content published in the previous hour was not appearing at all, including at major news publishers. Before attributing a same-day drop to your own actions, check whether anyone else noticed — Search Engine Roundtable is usually first.

5. Did the results page change?

Impressions flat, position flat, clicks down means the layout changed, not your ranking. An AI Overview arriving on a head query does this. So does a competitor winning a rich result you used to have.

6. Did demand change?

Impressions down with position flat means fewer people searched. Check the query trend. You cannot fix a demand drop with SEO.

7. Only now: was it an update?

If nothing above explains it, check whether a confirmed update overlaps the date. Note the shape: a core update drop is usually gradual across the rollout window and site-section-wide. A penalty is abrupt. A technical fault is abrupt and usually total.

Isolate before you conclude

Segment by page type, query type, country and device before deciding what happened. "The site is down 20%" is rarely true. "Blog posts in the US on mobile are down 60%, everything else flat" points at a cause; the aggregate hides it.

What trips people up

• Changing things during the diagnosis. Now you cannot attribute the recovery.
• Reading the aggregate.
• Believing the first plausible story. Confirm the size and shape match.

You have got this when

Someone reports a drop and your first three questions are about tracking, indexing and segmentation — not about the algorithm.