The do’s and don’ts of SEO: common mistakes to avoid

01 / Priorities

Which SEO mistakes should you fix first?

Fix anything that prevents people or search engines from reaching an important page before polishing that page’s copy. A missing product category, an accidental indexing block and an awkward title are different problems. They do not belong at the same priority just because an audit labels each one an error.

Use this triage order as a working recommendation, not an algorithm score. Check how many useful pages are affected, whether those pages generate qualified demand, and whether the issue began after a release. Save a baseline before changing anything.

02 / Targeting

Do match search intent. Don’t write to a keyword count.

Choose a page format that helps the person finish the task behind the search. Someone looking for common SEO mistakes needs explanations and checks. Someone comparing an SEO provider needs scope, proof, commercial fit and a way to speak with the team. Repeating the same phrase does not resolve that difference.

Review the queries that already produce impressions, then inspect current results in the market you serve. Compare guides with guides and service pages with service pages. Use the vocabulary your customers use without forcing it into every heading. Google identifies unnatural repetition as keyword stuffing.

• Do: give each important page a clear audience, question and next step.
• Don’t: create separate pages for every slight wording variation when the answer is the same.
• Verify: review the actual landing page for each query group, then ask whether it satisfies that visitor’s need.

03 / Access

Do separate crawling from indexing. Don’t confuse their controls.

Crawling is fetching a page; indexing is deciding whether to store and use it for search. A robots.txt restriction controls crawling. A noindex directive tells a supporting engine not to index a page, but the crawler must be able to access the directive to read it. Google documents this distinction in its robots rules.

If a page is unexpectedly absent, inspect the URL’s response, robots rules, meta robots, X-Robots-Tag header and canonical. Compare the live test with the indexed version in Search Console; they describe different moments. A staging noindex setting left on production is a different diagnosis from a page that Google crawled but chose not to index.

• Do: preserve intentional protections for private, account and test pages.
• Don’t: remove every exclusion or repeatedly submit URLs without understanding the reported reason.
• Verify: test representative URLs in every affected template and inspect the rendered content after release. A successful response alone does not guarantee indexing.

04 / URL recovery

Do recover useful missing pages. Don’t send every old URL home.

When a useful article disappears, first decide whether its content still belongs on the website. Restore it at its original address if it does. If a genuinely equivalent replacement exists, use a permanent server-side redirect to that page. If the content is gone and there is no suitable replacement, a real 404 or 410 is appropriate.

A homepage is rarely a replacement for a specific how-to article. Google warns that irrelevant migration redirects can confuse visitors and be treated as soft 404s. A large backlink count is not a reason to ignore the intent of the linked page.

• Do: compare the old URL inventory with the new site before a redesign launches.
• Don’t: assume a URL works because clicking the homepage navigation works.
• Verify: request the exact old address, follow the redirect if present, and check the final canonical, internal links and sitemap. Save this test with the release record.

05 / Canonicals

Do consolidate genuine duplicates. Don’t canonicalize different pages together.

Duplicate content is not automatically a spam violation. It can still confuse visitors and leave search engines choosing between competing versions. Google’s canonicalization guidance applies to duplicate or very similar pages; it is not permission to point every product or location page at a generic category.

Choose the preferred version, align its canonical with internal links and the sitemap, and redirect redundant versions when they no longer need to remain available. Distinct products or locations with distinct user needs deserve a separate decision. Do not use noindex merely to force Google to choose another canonical.

• Verify: inspect both the declared and Google-selected canonical after recrawling. A canonical tag is a signal, not a command.

06 / Page clarity

Do use descriptive titles and links. Don’t hide the page’s purpose.

A title and main heading need to describe the page visitors actually receive. Replace boilerplate that could fit any article with a specific description of the subject. Don’t repeat a long list of near-identical keywords or promise an answer the page does not deliver. Google can generate a different title link, so an exact display is not guaranteed.

Connect relevant guides, services and proof with meaningful links. Google recommends crawlable anchor elements with href attributes and concise, descriptive anchor text. A visual button without a navigable destination is not a substitute. A sitemap is useful, but it does not replace a helpful route through the website.

For example, a reader diagnosing overlapping pages can continue to on-page SEO support. A reader with widespread access problems can investigate technical SEO. The link belongs where that next step makes sense, not in every paragraph.

• Verify: read the title, main heading and anchor text without the surrounding design. Follow each link and confirm that the destination meets the expectation it sets.

07 / Editorial quality

Do add evidence and useful detail. Don’t publish a longer paraphrase.

A useful article helps readers make a decision or complete a task they could not complete before. Add the inspection steps, exceptions, examples or evidence that the answer requires. Google’s people-first content guidance specifically asks whether a page adds original value instead of simply rewriting other sources. It does not prescribe a preferred word count.

AI assistance does not remove the need to check factual claims, dates, citations and examples. Invented client results and screenshots are not evidence. Scaled pages produced primarily to manipulate rankings can violate Google’s spam policies regardless of how they were written.

• Do: identify who created the content and distinguish documented findings from recommendations.
• Don’t: change a date to imply that an unchanged article received a substantive update.
• Verify: open the cited sources, check each numerical claim and ask an appropriate subject-matter reviewer to test the instructions before publication.

08 / Authority

Do inspect backlink context. Don’t buy authority by the number.

Assess why a page links to you, where the link appears and whether it helps the referring site’s audience. A third-party domain score does not establish editorial endorsement. An unrelated comment on a strong domain is not equivalent to a relevant publication recommending your work.

Google classifies links created primarily to manipulate rankings as link spam. Paid advertising is a separate activity: paid placements need appropriate qualification, normally sponsored. Comments and forum links have their own ugc relationship. See Google’s outbound-link guidance.

Do not automatically disavow a list because an SEO tool labels it toxic. Google’s disavow guidance sets a narrower threshold involving substantial unnatural links and an actual or likely manual action. Review the situation before taking that step.

• Verify: sample referring pages and record placement, relevance, link attributes and whether the link still exists. Keep raw link totals separate from verified editorial references.

09 / Usability

Do test the mobile journey. Don’t optimize only a speed score.

A page can look polished on a desktop and still fail when someone opens a menu, reads a table or submits a form on a phone. Test the main journey at narrow widths, with a keyboard and with a slower connection. Check that content stays readable and that intrusive overlays do not hide the answer.

Use performance reports to find actual bottlenecks. Google uses Core Web Vitals in ranking, but good scores do not guarantee top positions. Improving an irrelevant page’s score does not make its answer relevant. Prioritize loading, responsiveness and layout problems that affect the visitor’s task.

• Verify: repeat the same journey after the change, record what improved, and check field performance where sufficient real-user data exists.

10 / AI search

Do clarify entities. Don’t treat schema as an AI-ranking switch.

Structured data gives explicit labels to information on a page, including people, organizations and articles. It can help Google interpret that information and establish eligibility for supported rich results. It does not replace accurate, visible content. Follow the structured-data documentation and keep the markup consistent with the page.

For Google’s AI features, there is no special schema requirement or mandatory short-block format. Google’s generative AI optimization guidance rejects prescribed content chunking and unnecessary AI text files as ranking tactics. Break up an article to help people follow it, not to satisfy an assumed hidden word limit.

Do not extend a statement about Google into a claim about every AI product. A ChatGPT mention, a citation, a recommendation and a visit are different observations. Record the question, date, answer and cited URL when testing visibility; a single response is not proof of stable rankings or business impact.

• Verify: validate the markup, compare it with visible facts, and report AI mentions separately from referrals and qualified enquiries. Our AI SEO service connects this work to the pages and evidence a team can actually improve.

11 / Business results

Do measure qualified demand. Don’t label every click a lead.

Use search reports to understand discovery and analytics to understand on-site behavior. Then use verified lead or transaction records to evaluate the commercial result. Google explains why Search Console clicks and Analytics sessions differ; they are not interchangeable measurements.

Choose the outcome that fits the business. For a service company, that might be qualified calls or accepted sales enquiries. For an online store, it might be completed purchases and recorded revenue. A submitted form is not automatically a qualified lead, and a page-view event is not a completed booking.

• Do: compare like-for-like periods, markets, page groups and definitions. Annotate releases, campaigns and tracking changes.
• Don’t: attribute every increase after an SEO change to that change when other factors moved too.
• Verify: send an authorized test enquiry, confirm where it lands and ensure your reports count it correctly. Exclude test activity from commercial reporting. Use SEO analytics and reporting to connect the records.

12 / Working checklist

Turn an SEO finding into a testable fix

Every accepted fix needs an affected URL, an owner and an observable acceptance check. Keep implementation evidence separate from the later search result. A developer can verify that a page now returns the right content; that does not establish when a search engine will recrawl it or how it will rank.

The example below is illustrative, not a client result. Copy the fields into your issue tracker and replace the example with your own evidence. If you cannot state what would prove the change worked, the task needs a clearer diagnosis before implementation.

For a broader starting checklist, read our SEO tips for Philippine websites. This guide focuses on diagnosing mistakes and validating repairs; the broader checklist helps plan an ongoing programme.