Search Engine Optimization: How to Use It on a Site That Already Exists
Using search engine optimization means running a short, repeatable loop: check what the site already tells search engines, remove what blocks indexing, improve one page at a time, and read the result every week. You do not need a new website, a course, or a hundred blog posts to start. You need an order of operations and a fixed weekly slot.
Six chapters, one weekly hour. Start at chapter 01, not at a keyword tool.
How do you actually use SEO on a website?
You use SEO by working in a fixed sequence: confirm the site can be crawled and indexed, find pages that already earn impressions but underperform, improve them one at a time, and review impressions, clicks, and inquiries every week. The routine matters more than any single tactic, because search results respond to accumulated changes.
Quick index: start where your problem is
Most teams do not need the whole manual at once. Find your situation in the table, read that chapter, and come back when the loop is running.
| If this is your situation | Start at | What you will leave with |
|---|---|---|
| A site that exists but gets no search traffic | Chapter 02 | A crawl and index check, plus the first three fixes |
| Traffic dropped and nobody knows why | Chapter 02, then 05 | A date-stamped diagnosis order instead of guesswork |
| Traffic arrives but no inquiries | Chapter 03 | Two rewritten pages with clearer answers and CTAs |
| No time, no owner, no routine | Chapter 04 | A one-hour weekly loop with four named actions |
| Reports look fine, decisions still feel random | Chapter 05 | An order for reading metrics and one monthly number |
| Two migrations, a redesign, or a manual action | Chapter 06 | A short list of triggers for outside help |
Collect the baseline before you touch anything
The first hour of SEO is not research. It is recording what the site already does, so that six weeks from now you can tell whether a change helped or whether the market moved without you.
Open Search Console and export ninety days of queries by page. List every URL in the sitemap. Note the five pages that already produce inquiries, even if they get little traffic. Those five pages are the most valuable thing on the site, and most teams cannot name them.
- Queries by page, last 90 days, exported to a sheet
- Every sitemap URL with its current impressions and clicks
- The five pages that produced a quote request or a direct message
- One screenshot of the homepage result for your brand name
- The date of the last deploy, migration, or URL change
That last item explains more traffic drops than any algorithm update. If a deploy happened three weeks before the decline, read chapter 02 before reading anything else.
Fix the floor: crawl, index, and speed
Before improving a single sentence, confirm the machine can reach the page. A blocked URL, a stray noindex, or a canonical pointing at the wrong version will make every later effort invisible.
Run the same five checks in the same order. They take an afternoon on a small site and a week on a large catalogue, and they remove the most common reasons a good page never appears.
- robots.txt allows the sections that should rank, and does not block CSS or JS
- The sitemap returns 200 and lists canonical URLs only
- Every important page has a self-referencing canonical
- No noindex tag survives from staging, and 404s return 404 rather than 200
- Redirects resolve in one hop, and the mobile page is the same page
Do not change URLs and rewrite page copy in the same week. If the number moves, you will not know which change caused it, and you will be unable to repeat the result.
If the site has thousands of filter URLs, this chapter is where crawl attention disappears. Reducing the URL set is usually more effective than adding server capacity.
Improve one page at a time
Pick a page that already earns impressions for a query it does not win. Position eight to twenty is the sweet spot: the page is eligible, the query is relevant, and the gap is small enough to close with real work.
Rewrite the opening so the answer arrives in the first sixty words. Add one table that a buyer would screenshot. Add two internal links from pages that already have traffic. Add a proof element: a measurement, a process photo, a tolerance, a lead time. Then stop and wait three weeks.
One page per week is a realistic pace for one person. Four to eight pages a month is a good month. Rewriting ten pages in a weekend usually means none of them were finished.
Where this connects to execution
Titles, headings, internal anchors, and structured data are page-level craft, and they are the difference between a page that answers a question and a page that ranks for it. If your team can write but cannot structure, On-Page SEO Services covers that layer without replacing your own expertise. If the basics of what search engines reward are still unclear, read Whats Is SEO? before rewriting another page.
Do not judge a rewritten page for two weeks. Early movement is noise, and reverting on day four is how teams end up with a site full of half-finished edits.
Chapter 04: the weekly routine
This is the part that decides whether any of the previous chapters compound. One hour, four actions, the same order every week. Put it in the calendar as a meeting with one named owner.
Monday: read the changes
Twenty minutes in Search Console. Note which pages gained or lost impressions and which queries are new. Do not act yet.
20 minTuesday: improve one page
Answer first, table second, two internal links, one proof element. Finish it or leave it out of the count.
40 minThursday: link one article
Find a page with traffic and add one contextual link to the page you improved. Ten links a month beats a site-wide block.
10 minFriday: log three numbers
Impressions, quote requests, pages improved. One line each. A log beats memory when the quarter ends.
5 minSkip a week and nothing breaks. Skip a month and the loop stops producing decisions, which is the real failure mode. The routine is intentionally small so that a busy week cannot erase it.
Read the numbers in the right order
Metrics are a sequence, not a dashboard. Read them in this order and each one tells you whether to look at the next.
| Metric | Where it lives | Question it answers |
|---|---|---|
| Impressions | Search Console, Performance | Is the page eligible and being seen at all? |
| Clicks and CTR | Search Console, Pages tab | Does the result earn the click once it appears? |
| Average position | Search Console, Queries tab | Is the page close enough that improvement is realistic? |
| Enquiries | Analytics or CRM | Is any of this turning into a conversation? |
If impressions are flat, the problem is eligibility, not the title. If impressions rise and clicks do not, the snippet is the problem. If clicks rise and inquiries do not, the page promises something the offer does not deliver.
Two cautions. Average position is an average across queries and locations, so it is a trend, not a rank. And a third-party rank tracker uses its own location and device, which is why it rarely matches the number in Search Console.
A rising session count with flat inquiries is not progress. Report one business number next to the traffic number, every month, or the routine will slowly optimize for readers who never buy.
When to bring in outside help
Most of this manual is designed to be run in-house. There are four moments where outside help saves more than it costs, and each one is a specific technical condition rather than a feeling that growth is slow.
Pages will not index. If important URLs sit in "discovered" or "crawled, not indexed" for more than a month, the cause is usually structural: crawl waste, rendering, or duplication. That is a diagnosis job, not a writing job.
Traffic fell sharply and the cause is unknown. A thirty per cent drop with no obvious deploy needs log files, coverage history, and a comparison against the previous quarter. Guessing at titles makes it worse.
The site is about to change platform. Migrations concentrate risk into two weeks. A review before the move is cheaper than a rescue after it.
Nobody owns the loop. If chapter 04 has been skipped for two months, the problem is capacity, not knowledge. Bringing in a team for the routine is a valid use of budget.
Outside help is not a substitute for the baseline in chapter 01. The best brief you can send includes the export, the five converting pages, and the date of the last change.
Send a message and we will reply with the six-chapter checklist and the weekly log template.
Questions readers ask before chapter 01
What should I do first on an existing site?
Export ninety days of queries from Search Console, confirm the sitemap and robots rules are clean, and list the five pages that already produce inquiries. Fix blocking issues before writing anything new.
How long does SEO take to work?
Page-level improvements often show movement in six to twelve weeks, while structural fixes can take a quarter to settle. Judge the consistency of the routine rather than a single weekly ranking.
Do I need to publish new content every week?
No. Improving one existing page per week usually beats publishing a new one, especially on a site with fewer than a hundred pages. New pages make sense when a query has no owner at all.
How many pages should I improve per month?
Four to eight is realistic for one person doing the work properly. Each page needs a rewritten answer, internal links, and a measurement window, and those three steps are the whole job.
Is SEO a one-time task?
No. The technical floor needs occasional repair and the query landscape shifts, so a short weekly loop matters more than a large annual project. The loop is also what makes the annual project worth doing.
Can I run this without an agency?
Yes, for most small and mid-sized sites. The manual assumes one person with an hour a week and access to Search Console. Bring in help when the escalation triggers in chapter 06 appear.
Set the routine once, then let it run
Send your domain and the last ninety days of Search Console data. We will return the baseline sheet, the first five pages to improve, and a weekly log your team can actually keep.