From SEO audit to useful pages: our step-by-step workflow

1. Record what you know—and what you cannot yet measure

Start with one row per important page. Save its URL, title, main heading, intended reader, next action, canonical URL and the date checked. Record the actual symptom: a vague title, repeated exercise, unsupported claim, broken download or confusing booking step. Keep the source beside the finding so another person can reproduce it.

If authorized data is available, export Search Console clicks, impressions, click-through rate and average position with dates and filters. Keep country, device, query and page definitions consistent. Use analytics or booking records for a separately defined business outcome, such as a qualified enquiry. Search clicks and confirmed appointments are different measurements.

In our audit, Ahrefs requests were blocked by plan access and the Search Console connector required a subscription. We recorded those gaps instead of treating unavailable data as zero. We could still inspect public content and delivery problems, but could not establish which query had lost traffic or which competitor had gained it.

Finish this step with: an evidence register and an explicit list of missing measurements. A sitemap entry proves that a URL is listed; a successful HTTP response proves that a server returned content. Neither proves that Google indexed the page.

2. Choose a visitor problem you can actually improve

Give each site one clear role. For a local coding site, a parent may need to understand the project, location and whether a class is confirmed. For an AI tutorial, the reader needs a task they can complete. For a service page, the reader needs to know the problem, scope and next step.

Prioritize a page when the problem is directly observable, the correction helps the intended reader, and you can verify completion. Fix false availability, inconsistent offers and broken paths before adding new articles. Then improve existing pages that answer a useful, specific question. Search demand and ranking opportunity should influence the next round when real data becomes available.

Finish this step with: a short change brief: “For this reader, improve this page so they can do this task; verify it using these checks.”

3. Replace vague search titles with specific descriptions

A slogan can work as supporting brand copy while leaving the search title unclear. Describe the service, subject or location a reader will find. Keep the visible heading and page content consistent with that promise. Do not add a city, service or qualification the business cannot support.

For example, Ecom Coach previously used “More from every customer journey.” The replacement names Shopify, conversion and retention. That makes the page's subject easier to understand; it does not demonstrate a click-through-rate improvement. Google may generate a different title link using other page signals.

Check it: inspect the public HTML title, main heading, description and canonical after deployment. Record the date. Later, compare relevant query performance and the displayed search result when data is available.

4. Make every tutorial produce a checked result

A list of tips becomes a tutorial when the reader can follow it from a starting input to a finished artifact. Use numbered actions, explain what success looks like, and show how to recover from a likely mistake. A graphic should reveal a decision, relationship or correction that is harder to follow in prose.

A worked example: tentative meeting notes

Our AI Teachers lesson supplies fictional notes, an exact prompt and a deliberately flawed output. One source says Friday might work, subject to availability. An incorrect draft turns that into a confirmed Friday action owned by Maya. The correction keeps the date unconfirmed and the owner unassigned unless another note establishes them.

1. Read the input. Mark tentative statements, missing owners and conflicting notes.
2. Make the first attempt. Use the supplied prompt and keep the output for comparison.
3. Check each claim. Match the action, owner and deadline to a source note.
4. Correct the output. Preserve uncertainty; do not fill gaps with plausible names or dates.
5. Try the changed input. Use the additional contradiction in the exercise and check whether the answer changes appropriately.

Try the complete meeting-notes tutorial, its practice PDF and checking worksheet. Other subjects need different artifacts: a small coding project, a printable gift planner, or a workshop evidence record.

Graphics check: give each diagram a text description, readable labels and a caption. Check it on a narrow screen. Use real screenshots only when they capture an actual tested process; label fictional examples and illustrations. A polished drawing cannot substitute for a tested result.

5. Resolve duplication by asking what job each page does

Compare the reader question and next action, not just matching phrases. Two pages about the same city can serve different needs: a general activity hub and a specific evening itinerary. We kept distinct West Palm Beach pages with their own canonicals and reciprocal navigation because their purposes differed.

1. Write the main question answered by each overlapping page.
2. If the questions differ, strengthen the distinction and link between the pages where useful.
3. If the questions and useful material substantially match, choose a primary page and combine the unique value.
4. Before retiring a URL, review its traffic, links and dependencies. Plan a redirect to the relevant replacement and update internal links and the sitemap.
5. Check the old and new URLs after release. Do not use canonical tags to paper over unrelated pages.

We also differentiated Wealth Experiment's documented workshop learning from AI Side Hustle's first-offer practice. The shared teaching structure could stay; the evidence, examples and reader outcome needed to be different.

6. Publish a bounded change and check the public result

Save the original and the intended replacement. Keep the change small enough to review and reverse. Use the site's existing publishing system, then confirm that the correct production site received the change. A merged change is an intermediate step, not the final proof.

• Content: one clear main heading; accurate dates, prices and availability; complete steps and an honest next action.
• Assets: diagrams render, downloads open, figures have descriptions, and worksheet arithmetic is correct.
• Discovery: the page has a relevant internal link, intended canonical and sitemap entry where appropriate.
• Behavior: check links and permitted test paths. For forms, distinguish validation failure, a successful submission and a confirmed business outcome.
• Layout: check narrow, medium and wide screens for horizontal overflow, clipped headings and unreadable diagrams.

Our October 29 follow-up recorded successful responses for 49 selected URLs and exact matches for all five replacement titles. A separate responsive pass covered 15 selected snapshots at three widths, giving 45 viewport checks. These were sampled release checks, not a full test of every interaction or proof of search indexing.

For an analytics example, use the GA4 lead-verification tutorial. Its instructions and worksheet do not mean your property or events have already been configured.

7. Measure outcomes and give the best page relevant exposure

Keep a change log and compare equivalent reporting periods after data has settled. Use the same filters and explain tracking changes, campaigns, seasonal demand and other releases. A 28-day comparison is a practical starting point, not a promise that rankings will change within 28 days.

Choose the metric that matches the page: relevant search clicks for discovery, successful practice completion for a tutorial, and qualified enquiries or confirmed bookings for a service journey. If the baseline is missing, say so and establish prospective measurement. Do not manufacture a before-and-after percentage.

Promotion is a separate task. Add a useful link from a related existing page, identify communities or partners that genuinely need the resource, and prepare a specific reason it helps them. Send outreach only with authorization. Track referrals and earned links separately; cross-linking an owned portfolio is not independent validation.

Our next evidence needs include real tutorial recordings, permissioned firsthand examples, genuine customer feedback and measured search outcomes. The implementation alone does not resolve those gaps.

8. Use the release checklist on one page

Download the editable CSV checklist No email or account required. Open it in your spreadsheet tool, duplicate the blank row for each page, and add evidence before marking a task verified.

1. Choose one important existing page and capture its current state.
2. Write one reader outcome and one observable problem.
3. Make the improvement, including a worked example or visual where it helps.
4. Check the published URL and record any remaining limitations.
5. Set a measurement date and decide whether to keep, revise or extend the change.

For the implementation decisions and commercial context, read Waymaker Agency's owned-portfolio case study. For a planning workspace, explore the Growth Decision Lab.

Sources and evidence limits

The examples and historical release counts come from our October 28–29, 2026 audit, implementation log and follow-up checks. They describe our own portfolio, not an independent client experiment. The graphics illustrate the method; no ranking, revenue or booking lift is claimed.

• Google: creating helpful, reliable, people-first content
• Google: title links in search results
• Google: consolidating duplicate URLs

Google guidance reviewed October 29, 2026. The workflow and examples are our editorial interpretation.