If Google reports spam or a security problem but you cannot find the page, start with the exact notice and its scope—not a list of low-traffic URLs. Open the relevant Search Console property directly, distinguish a manual action from a security issue, and use any examples to investigate the content or system producing the problem. Missing examples do not establish that the site is clean.
A notice may concern a group of pages, an injected script, user submissions or behavior that appears only under particular conditions. The investigation needs to connect the reported issue to actual content and responses. A low CTR, an unfamiliar incoming link or a red score in an SEO tool cannot make that connection on its own.
Confirm which kind of warning you have
Save the exact wording, property, issue type, date and affected URL patterns shown in the report. Check the real Search Console account rather than signing in through an unexpected email link. If you manage several domains or URL-prefix properties, confirm that the notice refers to the host and paths you are investigating.
| What you see | What to do first | What not to infer |
|---|---|---|
| A Manual Actions report entry | Expand the entry, record the affected scope and read the specific policy and repair instructions | That every page matching a displayed path pattern necessarily contains the violation |
| A Security Issues report entry or a browser security warning | Check the Security Issues report and investigate the stated harmful behavior or compromise | That a normal direct visit rules out conditional redirects or injected resources |
| Crawled – currently not indexed, a duplicate or noindex status | Investigate the page's indexing state, instructions and canonical selection | That Google has issued a manual spam action |
| A third-party spam/content score or an unsolicited alert | Verify the source, tool definition and actual page evidence | That the score is a Google enforcement record |
| Only a decline in clicks or rankings | Compare affected queries and pages, technical delivery and search conditions | That the decline proves hacking or a particular spam violation |
Google's Manual Actions documentation explains that a human reviewer identifies a policy violation and that the report can describe part of a site or the whole site. It also says that not every page matching a reported pattern is necessarily affected. Read the specific action rather than treating “spam” as one universal diagnosis.
The Security Issues documentation covers hacking and behavior that could harm visitors, including malware and deception. Its example URLs are samples, not a complete list; an issue can sometimes have no examples. These reports have different purposes and review routes. “No issues” in Manual Actions does not prove that the server is secure or that an automated ranking change occurred.
Expand the investigation beyond the sample URLs
Build a working inventory from sources you actually have: CMS records, sitemaps, crawl results, web-server and application logs, known external entry URLs, files and routing rules. Preserve the source of each URL. A page created outside the CMS may never appear in a CMS export, and a normal crawl may miss an injected page without links.
A combined inventory is still only the union of the sources checked. It cannot prove that no other URL exists. Record missing logs, restricted directories, unexamined subdomains and other gaps explicitly. If the evidence points to a shared generator or compromised account, investigate that mechanism as well as the known URLs.
- Group by how content is created. Compare templates, publishing accounts, imported feeds, comment systems, profile pages and upload handlers. Nearby URL paths do not necessarily share a cause, and one compromised component may affect several paths.
- Compare affected and apparently normal examples. Look at the delivered title, main content, links, redirects and loaded resources. Check the underlying files, database records and shared configuration when you have authorized access.
- Relate findings to changes. Review deployments, component updates, new accounts and scheduled tasks around the observed period. A nearby timestamp is a clue; it does not prove that a specific plugin or person caused the problem.
- Extend from verified patterns. If a template contains an unauthorized link, find where that template is used. If a process generates new URLs, investigate the process and its outputs rather than deleting only the first example.
For a fictional example, a report supplies one spam page under /community/. Investigation finds that an unauthorized account created both forum posts and public profile pages. Checking only the reported post would miss another content type. The useful next step is to inspect that account's activity and the affected publishing paths; it is not to declare every legitimate community page spam.
A site: search with a distinctive phrase may reveal additional public results, but it is not a complete server or index inventory. A suspicious-looking parameter, an old directory name or a legitimate medical term is also not proof of an intrusion. Inspect the actual response and its origin before removing content.
Check the conditions under which the problem appears
Ask what was different when the issue was observed: direct entry or a search result, mobile or desktop, logged in or signed out, a fresh session or an existing cookie. Conditional behavior can make the owner's usual browser look normal while other visitors receive a redirect or injected content.
For suspected infected pages, Google's security guidance describes using URL Inspection or non-rendering HTTP requests to examine content without directly executing page scripts. Record status, response headers, body and redirect destinations. Use the relevant observed conditions where possible; there is no need to follow a suspicious destination just to preserve its address.
A request with a copied user-agent or referrer can help test a condition, but it does not become a verified Googlebot request. Record what you actually tested. Compare Google's stored crawl information with any current live test, since the stored version may predate a cleanup.
Check logs at the locations configured for your server and hosting platform. There is no universal Apache log path for every WordPress or Nginx installation. Record time zones and retention gaps. Repeated login attempts, POST requests or 404s can be investigative leads, but do not establish successful compromise without supporting evidence.
Match the repair to the confirmed problem
Unauthorized pages, links or code
Preserve a dated copy of relevant files, database content, responses and logs before cleanup, while containing any ongoing harm. Mark an infected snapshot as evidence rather than a trusted restore point. Coordinate with the host or person responsible for the server if the problem crosses application boundaries.
Remove the confirmed injection and the mechanism that can restore it: compromised accounts, vulnerable components, unauthorized scheduled tasks or modified deployment material, as applicable. Replacing a visible link or disabling the newest plugin does not establish that the entry point is closed. Review affected access and sessions from a trustworthy environment, and verify compatible updates and restored files.
WordPress's hacked-site guidance stresses documenting symptoms, coordinating with the host, backups and investigating how access was obtained. Its hardening guide provides further context on access, updates and monitoring. Apply controls to the actual installation: an Apache example is not automatically valid Nginx configuration, and a blanket restriction can break legitimate uploads or application functions.
For the broader recovery sequence, see the hidden-link and compromised-site cleanup guide (Chinese). Check the original triggering conditions again after the repair, including relevant application and CDN caches. A clean scan is useful evidence within its coverage, not proof that every page and condition was examined.
User-submitted spam or misleading published content
Inspect the submission path named by the evidence: comments, forums, public profiles, forms, uploads or another feature. Remove confirmed abuse and adjust the responsible permissions, moderation or submission controls. An account without published posts is not automatically abusive, and a technical discussion containing a term used by spammers may still be legitimate.
For the site's own content, compare the page with the actual policy concern. Check copied material, unsupported claims, deceptive promises and whether pages provide a useful answer. Google's spam policies describe specific practices; they do not create a universal 5% keyword-density threshold or classify every AI-assisted draft as spam.
Use editorial readability and logic checks to improve meaning, not to obtain a supposed compliance score. If URLs repeat useful content, use the duplicate-content decision guide to distinguish canonical preferences, redirects, exclusion and crawl controls. Do not rewrite correct product measurements merely to lower a similarity score.
A link-related notice
Read the exact report category and affected scope. Incoming links, links published on your own pages and unauthorized code injected into the site are different situations. Record the actual URLs, placements and ownership information before deciding what requires investigation. An unfamiliar domain or a vendor's low score alone does not establish an intrusion or justify bulk removal.
Keep any service assessment separate from the report's stated issue. Guangsuan's current link-service explanation (Chinese) describes the company's product roles, quantities and layered approach; a third-party risk score does not independently verify its delivery or explain a specific Search Console notice.
Restore useful pages; remove genuine unwanted content correctly
When an existing legitimate page was altered, restore its accurate content at the appropriate URL and verify its delivered response, title, links and indexing instructions. When a hacker created a page with no legitimate replacement, remove its generating mechanism and check that the unwanted URL actually returns a suitable not-found response. Redirecting every spam URL to the homepage leaves the underlying distinction unresolved.
Age and zero reported clicks are not deletion rules. A correct older manual may still help customers, and a page with no visible clicks may reflect limited data, query demand or canonical reporting. For the measurement question, use the high-impression, low-click troubleshooting guide; CTR below 1% is not a Google spam verdict.
If harmful, hacker-created URLs need urgent hiding from Google Search, the Removals tool can provide a temporary measure. Check the exact URL or prefix carefully. The tool does not clean server files, revoke access or remove content from the internet, and it is not the ordinary way to tidy up every old 404.
Do not remove the whole site from search merely to deal with a few injected URLs, or use a temporary removal to choose your preferred canonical. Robots.txt blocks crawling rather than reliably removing indexed content. Private information needs real access control; search hiding and website access are separate decisions.
Request review with a record of work actually completed
Address the full affected issue before requesting review through the relevant report. If several issues are listed, check each one. For corrected public pages that remain online, verify that Google can reach and inspect the repaired content. For deliberately removed malicious pages, verify the removal instead. This is not an instruction to expose private administration or customer data.
A useful repair record contains the exact issue, affected scope, confirmed cause or remaining uncertainty, changes made and retest results. Here is a practical structure to complete with your own findings:
- Scope: which URL groups, templates or publishing mechanisms were checked, and how the search extended beyond the supplied examples.
- Repair: what was removed, restored or patched, and which affected permissions or processes were corrected.
- Verification: the URLs, access conditions and dates actually retested, plus the observed result.
- Follow-up: monitoring for recurrence and any investigation gaps that remain unresolved.
Do not claim a full-site audit when only a sample was checked, or include credentials and personal records in the explanation. Google says reviews can take days or weeks, with some cases taking longer, and asks owners not to resubmit while a request remains pending. A review decision is separate from later recrawling, indexing, rankings and business results.
If there is no relevant manual-action or security notice, investigate the actual symptom using Google's traffic-drop guidance rather than inventing a penalty to appeal. For colleagues working in Chinese, the notice-based spam-content investigation guide summarizes the same distinctions.