How we recovered 20,217 deindexed pages and grew MachineHub's organic traffic 34%

Introduction

What do you do when Google quietly stops indexing your best inventory (and nobody knows why)?

MachineHub hit its organic traffic peak in 2024. Then pages started dropping out of Google's index for no obvious reason, and traffic fell month after month.

The team hired an SEO consultant and followed the advice, but nothing turned it around. By the time they came to us, more than 74,000 pages were stuck in the "crawled, currently not indexed" bucket in Google Search Console (crawled by Google, but kept out of the index).

This was never a content problem, so we did not treat it like one. We ran a technical audit, traced the collapse to a set of self-inflicted crawl and canonical faults, and handed the engineering team a prioritized fix list plus three implementation guides.

Once those shipped, the index grew by 20,217 pages, "crawled, not indexed" dropped by 19,501, and organic traffic recovered 34% (with no new articles and no link building).

The challenge

MachineHub's value lives in its inventory. That means type pages (used CNC lathes, drill presses, welding positioners), brand pages (Haas, Mazak), and individual listing pages.

For that inventory to earn traffic, every one of those pages has to be indexed. They weren't.

Our Jan 2026 crawl (Screaming Frog plus the PageSpeed Insights API, across roughly 10,000 sampled pages) showed the mechanism.

The most valuable category surfaces were telling Google to throw them away:

•303 pages on 302 "temporary" redirects. 194 /types/ and 107 /brands/ URLs pointed to the homepage, so Google read the catalogue's key surfaces as gone.
•1,386 pages canonicalized to a different URL, including 37 collapsed into a single "unknown brand" bucket (a data quality leak that marked these pages as duplicates).
•144 pages returning 200 OK but marked non-indexable, canonicalized away to the generic /search page.
•527 pages served as JSON instead of HTML. The /search and /listings templates relied on client-side rendering, and that also gave 10-second First Contentful Paint on 84% of pages and 15-second Largest Contentful Paint on 99%.
•1,739 broken 404 pages with live internal links still pointing at them (wasted crawl budget).

Stacked together, these signals told Google that big parts of the catalogue were redirects, duplicates, or thin render-dependent shells.

That is exactly how a healthy marketplace quietly bleeds 20,000+ pages out of the index.

Our solution

Here's the strategy in one sentence: Stop MachineHub from de-indexing its own inventory, then re-prioritize and enrich those same pages so they get back into the index and rank for the "[machine] for sale" queries the marketplace exists to win.

We advised and specified the work. MachineHub's developers shipped it.

#1: Unblock the index with a technical audit

We documented every fault and ranked it for the dev team.

The list was clear. Remove or redirect the broken internal links, fix the content type and server-render the /search and /listings templates so they return real HTML, then clear the load-time bottlenecks.

The rendering fix was the keystone. It explained the JSON responses, the slow load times, and the missing index coverage all at once.

#2: Re-prioritize the crawl with a rebuilt sitemap system

We added a fresh-listings sitemap (everything published in the last 24 hours, priority 1.0, lastmod always set to today, rebuilt daily).

That gives Googlebot a small, high-signal file to come back to every day.

Then we rebuilt the type and brand sitemaps with decile-based priority. Each URL scores from 1.0 down to 0.1 by active-listing count, so crawl budget flows to the densest, most commercial categories first.

#3: Enrich pages with structured data for commercial rich results

We added schema on top of what was already there (Product, Offer, ItemList, CollectionPage, and a category-level AggregateOffer with lowPrice, highPrice, and offerCount).

This is the pattern that helps category pages win "used [machine] for sale" queries and show up in Image pack and merchant listings.

#4: Consolidate competing category architectures

We merged the legacy /categories/ pages into a single /search/type/ pattern with 301 redirects.

We swapped the combinatorial internal links for related machine-type links, so crawl equity spreads across the catalogue.

Then we rewrote the generic meta descriptions into brand-specific copy.

One canonical surface, backed by smarter internal links.

The results

Over the eleven weeks after the fixes shipped, the recovery showed up first in exactly the pages we had unblocked (type, brand, and listing pages):

•+20,217 pages indexed, from 42,901 to 63,118 in Google Search Console.
•-19,501 "crawled, currently not indexed", from 74,287 down to 54,786.
•+34% organic traffic, recovered from ~25.9k to ~34.8k monthly visits (GSC impressions roughly doubled).
•Top-3 rankings went from 5 to 103, with 438 more keywords ranking overall.
•282 brand-new ranking pages, led by 91 type pages and 57 brand pages (the same surfaces that had been redirected to the homepage).
•A commercial recovery. The returning footprint is dominated by "for sale" and "used" buyer queries, now showing up in Image pack and AI Overview results.

What this proves

The most common thing we hear from established sites is "we've tried SEO and it stopped working."

Often the problem isn't SEO. It's that the site is quietly working against itself. A blog-and-backlink approach assumes the pages are indexed in the first place.

When a large catalogue is losing coverage, no amount of content or links will fix it. You have to find what's telling Google to look away.

MachineHub's recovery shows that the highest-leverage SEO work is sometimes invisible. No new pages, no new links. We just removed the self-inflicted signals that were burying an entire inventory.

Find the root cause, hand engineering a clean spec, and the rankings you already earned come back.