11.8x More Pages Crawled After a Replatform Recovery

THE CHALLENGE:

A replatform launched without an SEO migration plan cost this luxury real estate network years of accumulated visibility in a single release.

The January 2025 migration moved the site to a new platform but carried none of its search infrastructure across. Canonical tags pointed to the wrong URLs. Redirects from the legacy URL structure were incomplete. The structured data that told search engines and language models what each listing actually was — property type, location, price, availability — was lost in the rebuild. The damage showed up in the ranking profile: by June 2025, non-brand terms accounted for only 77.5% of the site’s total ranking keywords, meaning brand searches were carrying an outsized share of a business that depends on being found by people who don’t yet know the brand.

The root cause was reachability. The new platform rendered much of its navigation as JavaScript buttons rather than links, so entire sections of inventory existed on the site but could not be traversed by a crawler or an AI fetcher. Sitemaps had drifted from best practice. Internal linking was thin. In June 2025, third-party crawlers could reach fewer than 20,000 pages of a property network many times that size. For an inventory business where listings turn over constantly, that compounds with every new property published into a structure machines cannot navigate.

THE STRATEGY:

Moving Traffic Media recovered the site by rebuilding how it is discovered, not by chasing back the rankings it lost — and that decision is why the results show up in AI answers.

The obvious response to a broken migration is to fix redirects, restore rankings, and close the incident. That approach returns traffic to the pages that already ranked. It does nothing for the pages a crawler still cannot reach.

MTM sequenced the work in phases from March 2025, prioritizing by risk rather than shipping one large release on a site with continuously changing inventory. Redirect and canonical integrity came first, because nothing downstream matters if a machine resolves the wrong URL. From there the program addressed three questions in order: can every page be reached, does every page explain what it is, and do search engines and language models find out when any of it changes.

Two decisions shaped the outcome. The site’s location-based property search pages — the state, city, and neighborhood browse templates of the form “homes-for-sale” — were treated as the primary discovery layer rather than as secondary navigation, with titles, meta descriptions, and headings generated dynamically across the entire location taxonomy. Those pages are the bridge between broad search intent and individual inventory, and at a property network’s scale they can only be optimized programmatically. And internal linking was treated as the primary lever rather than external authority, because a property network needs breadth of coverage across hundreds of thousands of listings far more than it needs depth of authority on a handful of pages.

DETAILS:

The recovery ran across four connected workstreams.

Restoring crawl and canonical integrity. MTM audited the legacy URL structure against the new platform, corrected canonical tags that were consolidating authority onto the wrong URLs, and repaired the redirect map so equity from the pre-migration site resolved to its correct destination. Real estate schema was rebuilt at the listing level, and Product and WebSite schema were enhanced across the site to re-establish what the site is, what it publishes, and how its content is organized.

Making pages reachable. The JavaScript framework’s use of buttons in place of links was the single largest discovery blocker, and remediating it opened pathways to inventory that crawlers and AI fetchers had been unable to traverse since the migration. XML and HTML sitemaps were rebuilt to best practice. Location-based property search pages — the state, city, and neighborhood browse templates that sit between the homepage and individual listings — were opened to indexing with dynamically generated titles, meta descriptions, and headings across the full location taxonomy, turning the site’s largest template into its primary discovery layer.

Connecting pages to each other. MTM expanded internal linking at inventory scale through related, nearby, and also-viewed property widgets, so every listing became an entry point to adjacent listings. Anchor text was remediated sitewide so links described their destinations specifically rather than generically. Internal linking from the homepage was expanded to push authority and crawl priority deeper into the property set.

Signaling changes to AI and search systems. The site implemented the IndexNow protocol so new and updated listings are submitted for discovery on publication rather than waiting to be recrawled – a material advantage where inventory turns over constantly. An llms.txt file was deployed to make the site’s structure and key content explicit to AI systems that consume it.

Each phase shipped and was measured before the next began, so regressions could be isolated to a single release rather than a batch of simultaneous changes.

THE RESULTS:

The recovery restored the site’s reachability first, and everything else followed from it.

Pages crawled by Ahrefs grew from 19,898 in June 2025 to 235,471 in June 2026 — 11.8 times more of the property network reachable to third-party crawlers than a year earlier. That figure is the clearest measure of what the replatform had actually broken and what the recovery fixed.

The ranking profile rebalanced over the same window. Non-brand keywords accounted for 77.5% of the site’s total ranking terms in June 2025, in the wake of the redesign; by June 2026 that share had recovered to 85.2%. The site was once again being found by people searching for properties rather than for the brand.

AI visibility followed the same curve. Monthly sessions from large language models reached 3.3K in June 2026, a 190% increase over January 2026. July sustained the gain, delivering more than 3,000 LLM sessions, a 129% increase year over year measured against an established base rather than a standing start. The June figure was the peak in the reported window; July settled slightly below it while remaining more than double the prior year.

Breadth grew alongside volume, and it grew for a traceable reason. The site averaged 60 distinct pages cited by large language models per month in June 2026, up 88% from November 2025. Pages that cannot be reached cannot be cited, and the citation footprint widened as the reachable footprint did.

The durable outcome is not any single number. It is that the discoverability layer the replatform destroyed now works well enough that search engines and AI systems can find, parse, and cite the depth of the inventory rather than a handful of top-level pages.