On a site we manage, the best-performing page was redirected to a new address during a rebuild. When Google finished switching to that address, the page's average position went from 28 to 56 in about two weeks. Nothing broke. The redirect returned the right status code, the destination was canonical and in the sitemap, and every page loaded fast. By any technical checklist the move was clean.
The page was a free generator tool built for one specific use. It had earned roughly a third of all the search impressions the site had ever received. It was folded into a new tools section, at a URL whose tool did a related but different job, under the same name. Nobody compared the two.
An SEO strategy compounds only when every page that earns demand keeps doing the job the searcher came for. Keywords tell you what people typed; intent tells you what they expected to find, and Google ranks on the second.
What follows is the strategy we now run for clients' SEO. It is less a checklist than an order of operations, and every step exists because skipping it has cost something measurable.
What actually happened, in numbers
We read Search Console in half-month blocks, filtered to every URL containing the tool's name, so the old address and its replacement are counted together. Through the summer, that cluster drew between 69 and 99 impressions per half-month at positions in the mid-to-high twenties. After Google adopted the new URL as canonical, impressions went 90, then 46, then 14, and the average position sat around 56.
The mechanism is simple once you see it. A redirect passes the old URL's signals to the new one, and Google then re-evaluates the new page against the queries those signals came with. The queries were things like "json prompt to video generator free" and "video to json prompt". The new page did not mention video once. Google did exactly what it should: it decided the page answered those searches badly and moved it down.
The fix was to restore the original job, as the default mode, on the new URL rather than moving the page again. Six weeks later its position was back to 25, its best level. Impressions were not: they were running at about 30% of the July peak. Position recovers faster than lost demand does, and it is too early to say whether the rest returns.
Step 1: Sort your queries by intent before you touch a page
Export every query Search Console has for the property and classify it with a handful of regular expressions: brand terms, tool or free-resource terms, buying terms ("agency", "services", "retainer"), comparison terms ("vs", "pricing", "fees") and noise. It takes an afternoon and it changes what you think the site is for.
When we ran this on that site's 86 queries, free-tool searches carried 37% of impressions, buying searches 27%, and brand searches 23%. The buying searches sat at an average position of 74. That one table showed where on-page work could pay off and where it could not, which is the most useful thing a strategy document can do.
Step 2: Give every intent exactly one page that does that job
For each intent group, name the single URL that should win it, and write that page's job in one line: what the visitor gets when they arrive. If two pages share a job they compete with each other. If no page has the job, you have found a content gap, and no amount of optimization elsewhere fills it.
The trade-off is breadth. One page per intent means saying no to the second article on a topic you already cover, even when the keyword tool shows a tempting variation.
Step 3: Run an intent-parity check before any redirect or URL change
This is the step that was skipped. Before you remove, redirect, rename or rewrite any URL with impressions in the last 90 days, pull its queries, then write the old page's job and the new page's job in one line each. If the lines differ, stop. A redirect to a page with a different job is not a redirect; it deletes the ranking.
When parity fails, the order of preference is: update the page in place, redirect to a page that already does the same job, or build that page first and redirect second. And never move a URL Google has only just adopted as canonical; the temptation to "fix" a drop by moving the page again is strong, and it resets the clock.
Step 4: Let Google tell you what to prune
Search Console's "Crawled – currently not indexed" list is Google telling you it read a page and decided it was not worth showing. On the same site that list held four generic posts written years ago, plus a few legacy URLs. We set the posts to noindex, kept their content for rewriting, and took them out of the sitemap. The crawled-not-indexed count fell from 8 to 2.
Pruning feels like losing pages. In practice it concentrates crawl attention and internal links on the pages that can rank, and it costs you nothing that was earning traffic.
Step 5: Decide which gaps are on-page and which are authority
This is the concession most SEO strategies avoid, so here it is plainly: a page sitting at position 74 for a buying search usually cannot be fixed with copy. The site's service pages already used their core technology term dozens of times. More than half of the first page for those buying searches was agency directories and "best agencies" lists, and the site had eight referring domains, with most of its links coming from one partner's footer.
For gaps like that, the work is getting listed where the directories already rank, earning links, and publishing evidence other sites want to cite. Rewriting the page again would feel productive and change nothing. Knowing which kind of gap you have is worth more than any single tactic.
Step 6: Declare one entity, and keep the schema honest
On the sites we build, we publish a plain Organization with one name, one headquarters address, and the markets served expressed through areaServed. We avoid LocalBusiness for companies that sell beyond their city, because it is ranked by proximity and ties the business to a local catchment. Each service page carries Service markup, and each article carries BlogPosting with a real author and real dates.
The cost of honesty here is giving up schema tricks that might win a rich result for a while. A clean entity record is worth more over the life of a domain than any of them.
Step 7: Link every page to its service and its proof
Every article should link to the service it supports and to something that proves it: a case study or a working tool. Anmol Sweets grew on organic search and content with zero ad spend, and we link to it from SEO writing because it is the evidence, not because it is a convenient anchor.
Run the same audit on your templates. A common culprit is a related-posts block that always shows the same three articles, which leaves older posts never linked from any other article.
Step 8: Measure per page, every half-month
Site-wide totals hide exactly this kind of failure. The site's overall impressions dipped and recovered several times that summer for ordinary reasons; only the per-page view showed that one cluster had lost most of its demand. Compare each page against its own previous half-month, flag any page down more than 40%, and filter by URL substring so moved pages are read together with their old addresses.
Where this stops applying
None of this creates demand. If nobody searches for what you sell, intent mapping just tells you so faster. It also does not replace authority: a site with a handful of referring domains will not outrank established directories on competitive buying terms, however good its pages are. And recovery is not symmetric. A ranking lost in two weeks took us six to win back, and the traffic is still only partly back.
Questions to answer before your next change
- Can you name, in one line each, the job of every page that earned impressions in the last 90 days?
- Does any planned redirect send a page's searchers to a page that does a different job?
- Which of your weak rankings are content gaps, and which are authority gaps you cannot write your way out of?
- Would a per-page, half-month report have caught your best page losing most of its impressions within two weeks?
If you want a second pair of eyes on a migration or a redirect plan before it ships, our SEO team will run the intent-parity check with you.