How to improve SEO on a site that already ranks

The only list that matters: positions 5 to 20

If your site ranks for nothing at all, this is the wrong page — start with step-by-step SEO for a site you already have. This one is for the other situation: you rank, just not high enough for anyone to click.

Positions 5 to 20 are where the return sits. Those pages have already persuaded Google they belong on the topic. They don't need to be invented, indexed or argued for. Getting one from 11 to 6 is a smaller job than getting a new page from nowhere to 20, and it pays sooner.

Here's the exact pull:

• Search Console → Performance → Search results. Set the range to the last three months, not 28 days. You want a stable average, not last week's noise.
• Turn on both Query and Page, then export to a sheet.
• Keep rows where average position falls between 5 and 20, and impressions clear a floor that means something for your business — maybe 100 a month for niche B2B, 1,000 for consumer.
• Sort by impressions descending. The top twenty rows are your next quarter of work. Everything below that is a distraction wearing a spreadsheet.

The eleven moves, ordered by effort

Roughly in order of cost. The cheap ones at the top are not the weak ones — moves 1 to 3 take an afternoon and frequently produce the first visible change.

Moves 1–3: the cheapest lift on the site is click-through

Ranking and getting clicked are different problems, and almost everyone only works on the first one. Sitting at position 6 with a click-through rate well below what that slot usually earns means you're leaving traffic on a ranking you've already paid for.

There's no universal CTR benchmark worth quoting — it swings wildly by query type, by how many ads sit above the results, and by whether an AI Overview is occupying the top of the page. So don't chase an industry number. Compare each page against its own past, and against your other pages at similar positions.

• Lead the title with the query as people type it. If they search 'seo cost in india', 'Pricing | Company' will never win the click, however good the page is.
• Add the differentiator in the remaining characters — a number, a price, a year, a city, a constraint. 'Nine tests' beats 'a complete guide'. Specific beats comprehensive on a results page.
• Write the meta description as ad copy. It's not a ranking factor. It is the only sentence of yours a searcher reads before deciding, and Google rewrites it when it's lazy.
• Answer the query inside the first 60 words, in plain sentences, before any preamble. This is what a featured snippet and an AI answer both lift, and it's what makes the page look decisive to a skimmer.

Moves 4–6: refresh, rewrite, or leave it alone

Pages rarely fall because Google punished them. They fall because someone published something better and yours stopped being updated. That's content decay, and it's the most predictable traffic loss in SEO — every page you own is depreciating right now.

The decision between a refresh and a rewrite is not a matter of taste, and there's a two-minute test for it. Open an incognito window, search the query, and look hard at the top five results.

1. Refresh when the top five are the same shape as your page — same format, same intent, same kind of answer — and yours is simply older. Update the numbers, the dates, the screenshots, the examples. Cut the paragraphs that have aged badly. This is an hour or two and it's the best hour on this list.
2. Expand when the top five all answer three sub-questions your page ignores. Add those sections properly, with your own view in them. Don't pad to a word count; word count isn't the mechanism, coverage is.
3. Rewrite when the SERP has changed shape underneath you — your page is an essay and the top five are all comparison tables, or yours is a product page and Google now shows guides. Adding 500 words to a page that answers the wrong question doesn't fix the question.

Moves 7–8: internal links from the pages Google already trusts

Your site already has authority. It's just distributed badly — usually piled onto the homepage and whatever post got shared once three years ago, then trickled evenly across the footer.

Find your strongest pages: the highest count of referring domains if you have a link tool, or the highest organic clicks in Search Console if you don't. Then link from those pages, inside body copy, to the striking-distance URLs from section one. Three to five links per target, from three to five different pages.

Then audit the anchors you already have. Forty internal links reading 'click here' pointing at your pricing page tell Google nothing about the page. Descriptive anchor text is free and it's read as a description of the destination. More on why this outperforms most outreach in internal linking beats link building.

• Link from contextual body copy, not from a sidebar block that appears on 400 pages.
• Vary the anchor text naturally across links. Identical exact-match anchors 30 times reads as a pattern, because it is one.
• Check the link actually renders in the HTML. A link that only appears after a JavaScript click event may never be seen.

Moves 9–10: pruning, the one that feels wrong

This is the move people refuse to make, because deleting pages feels like deleting work you paid for.

But every URL is a claim on Google's attention. Pages that have earned nothing in twelve months don't sit there neutrally. They blur what your site appears to be about, they consume crawl capacity on large sites, and now and then one of them outranks the page you actually wanted ranked.

Pruning is not a synonym for deleting. In order of preference:

1. Consolidate. Two pages chasing one query is the most common fixable problem on a mature blog. Merge the better material into the stronger URL, 301 the weaker one, update the internal links pointing at it.
2. Improve. If the topic genuinely matters to your business and the page is just bad, rewriting it is better than removing it.
3. Noindex. For pages that serve users but have no business in search — thank-you pages, thin tag archives, internal tools.
4. Remove. Only for pages that serve nobody: no impressions, no links, no conversions, no strategic reason. Redirect to the closest relevant page if one exists.

Move 11: proving which change actually did it

You're about to make ten kinds of change across twenty pages. Do them all in one week and you will learn precisely nothing about your own site, which is the expensive part — not the wasted month, the wasted lesson.

The discipline is dull and it's what separates people who improve their SEO from people who keep buying advice about it.

• One change type per cycle. Titles this month. Internal links next month. Refreshes the month after.
• Log it. One row per URL: date, what changed, target query. A sheet is fine. GA4 annotations are fine. Memory is not.
• Compare 28 days after to 28 days before, same page, same query filter, in Search Console. Not sitewide traffic — sitewide traffic moves for a dozen reasons that have nothing to do with you.
• Keep a control group. Leave five striking-distance pages untouched. If they rise as much as the ones you edited, the market moved and you didn't.
• Ignore the first fortnight. Google has to recrawl, and rankings wobble after any change. A drop on day three is noise, not a verdict.