Your SEO score dropped: a 15-minute triage checklist

A score that drops overnight is either a real problem that can cost traffic within days, or measurement noise that needs no action. The goal of the first fifteen minutes is to tell which, and if it is real, to find the change that caused it. This checklist works the same whether the drop came from a DeskSEO alert, a one-off check in the SEO checker, or a colleague asking why the number is red.

Minute 0–2: read what changed, not the total

The total score is a sum of 20 weighted criteria (the weights are public on How scoring works). A drop of 6 points tells you little until you know which line moved. Open the latest check result and compare it with the previous run:

• Which criteria lost points?
• Which issues are new, and at what severity — critical, warning or minor?
• Did the final homepage address change (a new redirect target, a different host)?

Then work through new issues in severity order. A single critical issue matters more than five minor ones, even if the minor ones cost more points in total.

Minute 2–5: the critical group first

These issues can remove a site from search results or stop crawling. If any of them is new, everything else waits.

Issue	What it usually means	First check
Homepage unreachable	DNS changed, certificate expired, server down, firewall blocking unknown visitors	Open the site on mobile data; dig +short example.com; check certificate expiry in the browser
5xx error	Application crash, database down, exhausted memory or workers	Application and web server error logs at the time of the check
Homepage not returning 200	403 from a WAF or bot protection, 404 from a wrong document root, 401 from a password left on	curl -sI https://example.com/ and look at the status line and server header
noindex on the homepage	A staging setting or a plugin switch shipped to production	Page source and x-robots-tag header
robots.txt blocks the site or homepage	Staging robots.txt deployed, or a new rule matching too much	Read /robots.txt line by line
Canonical points elsewhere	Template change, wrong base URL in settings after a migration	View source, search for rel="canonical"

For the indexing-related ones, noindex and robots.txt mistakes and the canonical guide go deeper.

A 403 that only the checker sees

If the site opens fine in your browser but the check reports 403 or "unreachable", the likely cause is a firewall, CDN bot rule or rate limit that treats unfamiliar user agents or data-centre IP ranges as hostile. That is worth taking seriously: the same rules can challenge or block search engine crawlers if they are configured broadly. Look at the WAF event log for blocked requests from the checker's user agent (DeskSEOBot) and, more importantly, for any blocks on Googlebot.

Minute 5–10: find the change

Real drops almost always follow a change. Ask, in this order:

1. Was there a deploy or a content-management release since the previous check? Look at the deploy log or the commit history around that time.
2. Did a plugin, theme or CMS core update run? Automatic updates are a frequent source of duplicated canonical tags, a second H1, or a changed robots meta setting.
3. Did anything change at the edge? New CDN rules, a WAF turned on, a certificate renewed with the wrong names, DNS moved to another provider.
4. Did hosting change? Server migration, PHP or runtime upgrade, a new caching layer. Migrations are where staging settings get copied over.

If nothing changed on your side, check whether the hosting provider had an incident, and whether the domain or certificate simply expired.

Minute 10–15: confirm before acting

• Run a second check. In the free checker, press "Check again" (a recent result is otherwise reused for a few minutes); in an account, run an on-demand check from the site page. If the issue is gone, it may have been a short outage — still worth a look at the logs, but not an emergency.
• Check Search Console. URL Inspection on the homepage shows what Google saw on its last crawl, and the Pages report shows whether "Excluded by 'noindex' tag", "Blocked by robots.txt" or "Server error (5xx)" counts are climbing. Google reports with a delay, so a clean report on the same day does not prove all is well.
• Write down what you found and when. If traffic dips a week later, the timeline saves an hour.

When a drop is just noise

Not every drop needs action. The most common source of noise is server response time. TTFB is measured with a single request from one location, and it is worth up to 6 points: full marks under 0.6 s, half between 0.6 and 1.5 s, none above. A server that sits around 0.6 s will flip between full and half marks from run to run, moving the total by about 3 points with no change on the site. The same goes for a sampled internal link that timed out once. See the TTFB guide for how to tell a slow server from a noisy measurement.

Other changes that cost points but are not urgent: a title that grew past 65 characters, a few new images without alt text, a missing og:image. Fix them in the next routine pass.

How DeskSEO alerts work

To avoid emailing about noise, scheduled monitoring in a DeskSEO account only sends an alert when a scheduled check finds one of these compared with the check before it:

• the score drops by 5 points or more;
• a new critical or warning issue appears (new minor issues show up in the check result but do not trigger an email);
• the homepage becomes unreachable, or becomes reachable again.

Each alert links to the check that triggered it, with the full list of issues and fixes. Checks you run by hand do not send emails — you are already looking at the result. The alert thresholds are deliberately conservative; a 3-point wobble every few days is expected, a 5-point step usually is not.

The checklist on one screen

1. Open the latest result; list new issues by severity.
2. Any critical issue? Check reachability, status code, noindex, robots.txt, canonical — in that order.
3. Identify the change: deploy, update, edge config, hosting.
4. Re-run the check to confirm; inspect the homepage in Search Console.
5. Only TTFB or a single link moved? Note it and move on.
6. Fix, re-check, and record what happened.