Why is my organic traffic dropping? Find the cause in Search Console
Website traffic dropped suddenly? Diagnose it in Search Console step by step: algorithm update, deindexing, redirects, noindex, seasonality or tracking.
If you are asking "why is my organic traffic dropping" after a sudden fall, the cause is almost always one of seven things: a Google algorithm update, pages falling out of the index, a migration or redirect that broke, a robots.txt rule or noindex tag, seasonal demand, a change on the results page (an AI Overview or a lost rich result), or a tracking problem that only looks like a drop. Search Console can tell them apart in about ten minutes if you check in the right order: confirm the drop is real, find the date, find which pages and queries lost clicks, then inspect those pages. Prompts for Claude or ChatGPT below.
Sudden drop or slow decline
Decide this first. A sudden drop is a cliff: clicks fall by a third or more over a few days and stay down. Something happened on a specific date, and your job is to find what. A slow decline is a slope: a page loses a few clicks a week for months while newer, fresher results overtake it. Nothing broke. That is content decay, and the fix is a refresh list, not an investigation.
This post covers the cliff: the case where your website traffic dropped suddenly. If the chart in step 2 shows a slope, stop here and use the content decay prompt instead.
Why is my organic traffic dropping: the seven usual causes
| Cause | What the data looks like | Tool that confirms it |
|---|---|---|
| Algorithm update | Positions worse across many pages, starting within the dates of a ranking update | search_insights movers, plus Google's Search Status Dashboard |
| Deindexing | Some pages go to zero clicks and zero impressions; the rest are fine | inspect_url |
| Migration or redirect break | Old URLs lose everything, new URLs gain little or nothing | check_redirects, inspect_url |
| robots.txt or noindex | A section or the whole site drops on the day of a deploy | check_robots, inspect_url |
| Seasonality | Impressions fall with position stable; the same dip shows a year earlier | search_analytics on last year's dates |
| SERP feature change | Position and impressions stable, clicks and CTR fall | search_analytics by query |
| Tracking issue | Analytics shows a drop, Search Console clicks do not | search_analytics against your analytics tool |
Google's own guide to debugging Search traffic drops adds security issues and manual actions. The API does not expose those, so check both reports in the Search Console interface.
Step 1: confirm the drop is real
Two checks before any analysis.
Is it Search Console or only analytics? Search Console counts clicks on Google's side; analytics counts visits that ran its tag. If GA4 says organic traffic halved but Search Console clicks are flat, nothing happened in search: a consent banner change, a broken tag, a new cookie setup or a referrer change moved your numbers. Common after a redesign.
Is the window complete? Search Console data trails by about two days, and the last day or two are partial. A "drop" on yesterday's data is usually just missing data. search_analytics returns final data by default for that reason.
Then read your notes. A common way to use the server is to save site changes as notes with add_note ("Moved /blog to /articles with 301s on 2026-08-12"). list_notes brings them back at the start of a diagnosis and often the answer is in there. Google's own chart annotations in Search Console do a similar job in the interface; the Search Console annotations post covers both and when each one helps.
The pattern of the four numbers already narrows it. Clicks and impressions down with position worse: a ranking loss (update, technical problem). Impressions down with position stable: demand or indexing. Clicks down with impressions and position stable: the results page changed.
Step 2: find the date
The start date is the best clue you will get. Line it up against your deploys, your notes and the ranking updates on the Search Status Dashboard. Cliff the day after a release: look at the release. Cliff mid-way through a core update and you shipped nothing: probably the update.
A mobile-only drop points to a mobile template or rendering problem. A one-country drop points to hreflang or a regional server issue. A drop spread evenly everywhere points to the site or the algorithm. Discover (type: discover) is reported separately from web search and is far more volatile, so check which one fell.
Step 3: find what lost the clicks
Run search_insights with the movers recipe. It compares the current window with the previous one of the same length and returns four lists: decliners (with clicks, impressions and position for both periods), risers, new pages and lost pages. The lost list matters most in a sudden drop: pages that had clicks before and none now usually were deindexed, redirected or blocked.
Fifty pages each slightly worse is an update. One folder at zero is a technical break.
Step 4: inspect the pages that dropped
For pages that lost everything, inspect_url asks Google's URL Inspection API what it knows: coverage state, indexing state, robots.txt state, page fetch result, last crawl date, and both the declared and the Google-selected canonical. That one call separates most technical causes:
- "Excluded by 'noindex' tag" or an indexing state of blocked by meta tag or HTTP header: someone shipped a noindex.
- Robots.txt state blocked: confirm with
check_robots, which fetches the live file and shows the exact rule that matched. - "Page with redirect" on a URL that should not redirect, or a new URL that is "Discovered" but not indexed weeks after a move: run
check_redirectson the old URL and follow every hop. The site migration checks use case has the full sweep. - Google-selected canonical differs from yours: Google is folding the page into another URL, often a parameter or duplicate variant.
- Indexed, crawled recently, fetch OK: technically fine. Go back to the position data.
For hundreds of URLs, Index Watch on the paid plans inspects your sitemap URLs in the background and lists pages that left the index, with reasons. See the index coverage audit.
Step 5: rule out seasonality and the results page
Seasonality. Compare the same weeks last year. The compare option only looks at the immediately preceding period, so ask for two explicit date ranges with search_analytics. Search Console keeps about 16 months, so this works for the most recent months only. If last November dipped the same way, nothing is wrong.
A results page change. Position and impressions stable, clicks down: something now sits above you or answers the question on the page. An AI Overview arriving on your main queries, a lost FAQ or review snippet, or Google rewriting your title all look like this. Search Console does not report AI Overviews per query, so search your top losing queries yourself. Snippet work is covered in low-CTR pages.
What to do once you know
- Technical break (noindex, robots.txt, redirects, canonicals): fix it today. These recover once Google recrawls, usually within days to a few weeks.
- Algorithm update: change nothing during the rollout. Afterwards, compare your losing pages with what now outranks them. Google says recovery can take months.
- Seasonality: nothing to fix. Note it so next year's diagnosis is faster.
- Results page change: work on titles, descriptions and structured data, or accept lower CTR on queries an AI answer now satisfies.
- Tracking: fix the tag, not the site.
Whatever it was, save a dated note with add_note so the next diagnosis starts with it.
Limits
- Read-only. The assistant reads Search Console data and inspects URLs. It cannot remove a noindex, fix a redirect or request reindexing; those are yours to do on the site and in the Search Console interface.
- No manual actions or security report. Google does not expose them in the API. Check them in Search Console directly.
- Anonymized queries. Rare queries are withheld, so query-level totals can fall short of page-level totals.
- Data lag and history. About two days of delay, about 16 months of history.
- Quota. A full diagnosis like the one above is roughly 8 to 12 data pulls (
search_analytics,search_insights,search_chartand eachinspect_urlcount; notes and thecheck_*tools are free). The Free plan allows 15 a day and 40 a month on one property, so one diagnosis fits, two in a day may not. Solo removes the caps for 4.99 € a month. Details on the pricing page.
Not connected yet: the Claude setup and ChatGPT setup guides take a few minutes.
Ready to put your search data in your AI?
Set up the Search Console MCP, freeFree plan included. See pricing