How Do I Know If My Website Has SEO Problems?

Start with the symptom

Most people ask this question because something feels off. The site is live, the pages are written, and nothing is happening. That feeling is often correct. The next step is to work out what exactly is wrong.

Four symptoms are worth taking seriously, and each points to a different report:

• Your pages never rank for anything. Not ranking low. Not ranking at all, for any search. That points to an indexing or quality problem, and we work through every cause in why your website is not ranking on Google.
• Traffic stays flat while you keep publishing. New pages should move something. If they do not, Google is either not keeping them or not showing them.
• Your indexed page count has stopped growing. If you keep publishing and the number stays flat, Google is declining your new pages. That is different from Google not finding them.
• Almost all your traffic lands on two or three pages. That usually means the rest of the site is indexed but never shown.

None of these is a full diagnosis on its own. Each one tells you which report to open first.

The six reports to check

Google Search Console has six reports that matter for this question. Each answers something different. Most problems show up as a mismatch between two reports, so check all six before deciding what is wrong.

1. Pages: how many pages Google kept

The Page indexing report splits every URL Google knows about into indexed and not indexed, and groups the not-indexed ones by reason. The reasons are the useful part:

• Crawled – currently not indexed. Google read the page and decided not to keep it. A few is normal. Hundreds is a content quality problem.
• Discovered – currently not indexed. Google knows the URL exists but has not visited it yet. On a big site this is a crawl budget issue. On a small site it usually means weak internal linking.
• Duplicate without user-selected canonical. Google found several near-identical pages and picked one itself. Your canonical tags are missing or being ignored.
• Excluded by "noindex" tag. Sometimes intentional. Check this row after every release, because a noindex tag meant for a test site can ship to the live site by mistake.
• Soft 404. The page loaded but looked empty to Google. Common on category pages and search pages with no results.

Our guide to the page indexing report explains every reason row. To test one URL right now, the free page indexability checker reads its robots directives, canonical, and status code in one pass.

2. Sitemaps: is Google reading your sitemap?

The Sitemaps report is small, but people rarely open it. Check three things:

• Every sitemap you have is listed. Sites that changed CMS or added a blog often have a sitemap that exists but was never submitted.
• Status says Success, and the last read date is recent. A sitemap Google last read five months ago is not helping it find anything you published since.
• The discovered URL count looks right. A sitemap with 412 URLs on a site with 1,800 pages means most of the site depends on internal links to be found.

If the count looks wrong, check the file itself before blaming Google. The free XML sitemap validator tells you whether the syntax, size, or URL format is the reason it is being ignored.

3. Performance: what your pages earn

The Performance report is where most people start and stop, usually by looking at the clicks line. Ask it four questions instead:

• Did clicks and impressions move together? Both down means you lost visibility. Impressions flat with clicks down means people see your result but do not click it. That is a title or description problem.
• Did the drop happen on one day? Switch the chart to daily. A one-day drop has one cause, like an update or a site change. A slow slide usually means aging content or stronger competitors.
• Which pages moved? Open the Pages tab and compare against the previous period. A site-wide worry usually turns into a list of five or six URLs.
• Which queries moved with them? Filter queries by a topic word to see whether you lost one term or a whole topic. Our free GSC regex generator builds those filters for you.

4. Manual Actions: the two-second check

The Manual Actions report either says "No issues detected" or it lists a penalty. Check it first, because a penalty changes what everything else means.

• Check Security Issues in the same visit. Hacked content and malware warnings push traffic down in ways no on-page work can fix.
• A clean report means no penalty. If this report is empty, a ranking drop was not a penalty. It came from an algorithm change or from your site.

5. Core Web Vitals: speed problems

The Core Web Vitals report uses real visitor data, so it needs enough traffic to show anything. A small site with no data here has not failed. It just does not have enough visitors to measure yet.

• Read it by template, not by URL. The report groups similar pages. A Poor group almost always maps to one page layout, not to random pages. Our guide to fixing Core Web Vitals walks through the fixes metric by metric.
• Fix Poor, do not chase Needs improvement. A Poor group on important pages is worth real work. Polishing a Needs-improvement group while pages sit unindexed is not.
• Speed rarely explains a big drop. It decides close contests between similar pages. It does not move a page from position 4 to position 40.

6. Links: what points at your key pages

The Links report lists your most linked pages, the sites linking to you, and your top internal links. The internal list is the more useful half:

• Your important pages should be near the top of internal links. If the homepage and blog index take almost every internal link, the pages that make you money are not getting enough of them.
• A page with no internal links will not appear here at all. That absence is how you spot an orphan page.

The four-number check

If you only run one check, run this one. It uses two reports, takes about two minutes, and tells a technical problem apart from a content problem better than any tool score.

1. Submitted. Open Sitemaps and note the discovered URL count.
2. Indexed. Open Pages and note the indexed number.
3. Shown. Open Performance, switch to the Pages tab, set the range to 28 days, and count the rows. That is how many pages got at least one impression.
4. Clicked. Sort the same table by clicks and count the rows above zero.

Now read the gaps. A big drop between submitted and indexed is a crawling, duplicate, or quality issue, and the Pages report already told you which. A big drop between indexed and shown matters most: Google kept those pages but never found a search worth showing them for.

Run the whole check with one prompt

Everything above is a series of clicks: six reports, several tabs, two date ranges, and comparisons you hold in your head. It works, but it is slow, so most people do not do it every month.

CrawlRaven connects to Search Console and GA4 directly and serves that data over an MCP server, so an AI assistant can read your real numbers. Instead of clicking through six reports, you ask one question and get one answer with the data behind it.

Connecting the CrawlRaven MCP server

There is no API key. The server uses OAuth, so setup is a URL and a browser sign-in:

1. Connect Search Console inside CrawlRaven and let the first crawl finish, so there is data to read.
2. Add the server in your client. On Claude.ai and Claude Desktop, go to Customize > Connectors, add a custom connector, and paste https://mcp.crawlraven.com/mcp. ChatGPT needs developer mode turned on first. Claude Code, Codex, and Cursor take the same URL from the command line.
3. Approve the connection in the browser tab that opens, and let the tab return to the client on its own.

The client-by-client setup guide covers all six clients and the four errors behind most failed connections. The MCP documentation lists every tool, parameter, and scope.

The four prompts to use

Replace the bracketed site with your own. The first prompt is the one to save. It runs the whole six-report check in one pass.

Using CrawlRaven for [your site]: compare the last 28 days with the previous period, tell me whether clicks fell faster than impressions, list the five pages that lost the most clicks and the queries they lost, check the timeline for anything that lines up with the drop, and finish with the three highest-impact opportunities.

For [your site] over the last 28 days, list every page that earned at least one impression and count them. Then tell me what share of total clicks the top ten pages take, and which pages have impressions but no clicks at all.

Walk through daily clicks for [your site] over the last 90 days, point out the sharpest drops, and for each one check the timeline for a Google update or a note near that date.

Find pages on [your site] with real impressions sitting between position 8 and 20, show the queries driving them, and rank them by how many clicks one position of improvement would be worth.

The last prompt turns a diagnosis into a task list. A page at position 11 with 4,000 impressions is a much better opportunity than a page at position 60, and a report sorted by clicks hides that difference.

What the prompts cannot answer

To be clear about the limits: the MCP tools cover search performance, pages, queries, the timeline, target keywords, and ranked opportunities. They do not expose the indexing reason rows, the sitemap status, manual actions, or the Core Web Vitals groups.

• Indexing and canonical problems come from the Pages report and from CrawlRaven's own crawl, which checks indexability, canonicals, redirects, and status codes across the whole site.
• Manual actions and security issues take two seconds to check in Search Console. Keep them in your monthly routine whatever tools you use.
• Core Web Vitals data needs real traffic to exist. No tool can produce it for a site that has none.

What Search Console cannot tell you

Search Console records what happened. It keeps no record of what you changed, what your competitors published, or what your pages looked like before a drop. That missing context is why two people can read the same report and disagree about the cause.

CrawlRaven closes that gap by joining four datasets that normally live in different tools:

• Search Console performance, so every finding is sized by the traffic actually at stake.
• GA4 engagement, so a page that ranks but loses visitors right away is not counted as a win.
• Your imported Ahrefs or Semrush keyword lists, so what you planned to rank for sits next to what you do rank for.
• A 200-point crawl, which checks indexability, canonicals, redirects, and site structure, with every issue ranked by likely impact on rankings.

The output is one ordered plan: what to write, what to update, what to fix first. Prefer doing it by hand? The technical SEO audit checklist covers the same ground free, and the easiest way to audit your website compares every path. Pricing is a one-time lifetime licence, listed on the pricing page.

When your SEO is actually fine

Sometimes every report is clean. Pages are indexed, impressions track the index count, no manual action, vitals fine, links reasonable, and traffic is still disappointing. At that point SEO is not the problem, and another audit will not help.

Two explanations cover almost all of these cases:

• There is no demand. The searches you rank for happen a few dozen times a month. Position 1 on a term nobody types is worth less than position 20 on a term people do.
• People see your listing but pick someone else. Impressions are healthy, position is decent, and the click-through rate says your title and description are losing the choice.

Both look like SEO problems from the inside, but they are content and positioning problems. Another crawl will not fix them.

The 20-minute version

If you do nothing else this month, do this, in this order. The order matters because a finding in step one can make the later steps unnecessary.

1. Manual Actions and Security Issues. Two seconds. A finding here outranks everything below.
2. Pages report. Note the indexed count and the top three not-indexed reasons by volume.
3. Sitemaps. Confirm every sitemap is listed, marked Success, and read recently.
4. Performance, 28 days versus previous. Clicks against impressions, then the Pages tab for the URLs that moved.
5. The four-number check. Indexed pages against pages with at least one impression. Write the ratio down so next month has a baseline.
6. Run the monthly sweep prompt. Once connected, it does steps four and five in one pass and adds the timeline and ranked opportunities you would not assemble by hand.