How TrustList came back online and then rebuilt its own catalogue in 2026

How TrustList came back online and then rebuilt its own catalogue in 2026

TrustList came back this year after about a year offline, and it came back as an ordinary directory: a very large catalogue, much of it imported years ago, much of it thin. Around 17 September 2026, 24,350 live listings, about a quarter, had fewer than 50 words of profile text. On 25 September Google AdSense refused the site for low-value content.

We did not argue. We read our own catalogue the way a reviewer would, and set about turning it into a curated one. By the afternoon of 29 September, 29,722 of 72,564 live company listings had been rewritten to a single standard, with about 19,000 more queued to publish that day. On 28 September a /best page existed only where a published ranking backed it, about 12,000 of them, where a few days earlier about 18,000 had sat in the sitemap.

This article is the overview of that change, from ordinary to curated. Before the shutdown the site had steady traffic that we had never tried to monetise; when it returned, the numbers were small, and we treat them as a recovery baseline rather than a measure of demand.

The article covers what we found and what followed. It covers what we found, the rules we set ourselves, the work we have done since, and what is still unfinished. Other pieces in this series go deeper on single topics, and we point to them at the end.

Where we started: a site that came back different

The public directory that returned in 2026 was not the one that went away. It had been rebuilt underneath, and it covered more than software and IT services. Agencies, recruiters, accountants, outsourcers and funders now sat alongside the software listings.

The old catalogue came with it. Much of it had been imported years earlier, and a lot of it was never written by us.

Our search records show what the gap did. The site had earned a handful of clicks a day in 2025, then collapsed. Through the first half of 2026, during the rebuild, it had almost no visibility at all. Since August it has been shown to far more people than before, but at a poor average position. We wrote up those numbers in our earlier piece on search traffic, so we will not repeat them here.

The point for this article is simple. Low clicks today say little about whether people want a directory like this. They say the site is starting again, and that Google is deciding how far to trust it. Google's own guidance on creating helpful content asks whether a page was made for people and whether it adds something a reader cannot find elsewhere. A returning site with a large, mixed catalogue has to be able to answer yes to that. Ours could not, for a large share of its pages.

What the AdSense feedback said, and what we counted

The review told us the site did not show ads because of "low-value content". It also asked for content that is authentic and high quality, for signs of ongoing curation and maintenance, and for genuine user interest. The programme's own rules are public: see AdSense programme policies.

We counted what we were asking Google to judge, using the sitemap as it stood on 25 September 2026:

• about 99,600 listing pages, a quarter of which said almost nothing and most of the rest of which was a business's own website copy;
• about 18,300 ranking pages under /best, assembled from those listings, which had drawn some 255,000 impressions and 67 clicks at an average position around 70;
• 1,000 internal search-result pages that held no listings in their HTML at all;
• roughly 960 other pages, which were the editorial part: articles, checklists and the pages that explain how we work.

At the thin end, our records from 17 September showed 24,350 live listings, about a quarter, with fewer than fifty words of profile text. More than 12,000 companies had no text at all.

We also found housekeeping problems. Search Console still held 142 sitemap submissions from 2023. Our server already redirected all of them to the current sitemap, so they did little harm, but they were not what "structural maintenance" looks like.

None of this was hidden. It was our catalogue, and it was ours to fix.

The rule we set: rebuild it, retire it or remove it

On 25 September the owner set a rule for thin content. In short: improve what is thin if we can find sources for it, and remove what is gone or cannot be sourced. New high-value content gets its own sitemap.

That turned every thin listing into one of three outcomes:

What we find	What happens	What the public sees
The business is live and says what it does	We rebuild the profile in our own words	A real profile, offered in a curated sitemap
The business says it has closed, or its product has ended	We retire the listing	A page marked so search engines skip it, and removal from every ranking
Nothing can be found to write from	We remove it	The page goes away

There is a subtlety we kept. A dead website alone is not proof that a business has closed, so retirement needs the business's own statement. But a thin listing with a dead website and nothing else describing it has no sources to write from, so it falls under removal rather than retirement.

Two supporting changes made the rule visible to Google. We started marking rebuilt listings with the date they were rewritten. And we published a second sitemap containing only what we stand behind: rebuilt listings, our editorial articles, the topic rankings, the checklists and the pages that explain our process. The number of those pages Google actually indexes is now the measure of this programme.

Keeping thin pages out of search while the work happens

At sixty rewrites a batch, the first queue would have taken over a year. Every thin page would have stayed in Google in the meantime.

So we changed what we offer. Thin company and product pages stay up for visitors, but they now tell search engines not to index them, and they are left out of the main sitemap. The test is worked out from the text on each request, so a listing is indexable again the moment we rewrite it. Nothing is deleted by this step.

We built in exceptions so that we did not damage anything that mattered. A listing with a review, a claimed listing (where the business has taken ownership of its own space) and an IT service provider are not treated as thin whatever the text says. The last group is held out because of a search experiment we are running until 13 October. We did not want to de-index half of a test.

We also dealt with rankings. About 18,000 ranking pages were in the sitemap, many built from import tags and full of thin listings. A ranking is now offered to search only when at least five of its listings have a real profile and its subject is a real subject. On a local test copy, about 12,000 of roughly 18,700 rankings stayed. The 1,000 internal search-result pages are no longer indexed either, because Google asks sites not to index internal search results.

Cleaning out spam, junk and self-promotion

Thin was the first problem. The second, found on 26 September, was worse.

We ran a full-text scan of every live listing, 89,840 of them, for signs of spam: adult and escort spam, casino and slot spam in several languages, pills, payday loans, love spells, essay mills, crypto "recovery" services, parked and for-sale domains, placeholder and error pages, and legal-page text stored as a description. It flagged 1,719 listings. The scan did not decide anything. Six reviewers read every flagged listing in context against a written brief.

The verdicts:

• 164 listings were wholly spam and 24 partly spam. The usual cause was a domain that had lapsed and been taken over, so the text on our page described something entirely different from the business we had listed. A recruitment agency's listing carried adult content. A venture-capital firm's carried slot-machine lists.
• 12 were illicit businesses in their own right, mostly essay-writing mills, plus a crypto recovery service and an astrologer. Nine of them had claimed their own listings.
• 432 were junk: no description at all, just parked domains with asking prices, error messages, placeholder text, unedited website templates, or a privacy policy standing in for a description.
• 1,087 were false positives, such as iGaming software vendors and real firms writing in their own language. They stayed.

We removed 176 wholly spam or illicit listings. The partly spam ones are real businesses with injected text, so we held them back to be rebuilt instead. We also archived 52 guest posts, 51 of which were promotional: backlinks to the author's firm, "hire us" pitches and press releases. One article labelled as editorial turned out to be an unfinished generator stub whose text told readers to set a key for a fully written article. That went too.

We found a bug in our own code as well. Blocked reviews were not being hidden, because a filter used an operator our system did not recognise. We fixed that and hid seven reviews, among them a funds-recovery scam pitch and two businesses' own marketing copy posted as reviews of themselves.

Finally, we changed where adverts may run. The ad tag now loads only on pages that the curated sitemap would offer: never on a page we have marked not-for-indexing, and never on a listing or ranking still built from the old catalogue. We checked this live after the change. The rules we apply to guest posts and listings are set out in our listing guidelines, and a companion article in this series covers them in detail.

The fast pass, and what we parked

The owner then asked for a quicker route: write a short profile from each thin listing's own website, and record the contact details the business itself publishes.

1. We read each listed website once. Sites behind bot walls were never bypassed.
2. AI-assisted writers took 30 listings at a time under a written brief: third person, roughly 110 to 180 words, facts only from the page, no hype, no prices, no contact details in the text.
3. Automated checks refused any draft with first person, hype words, prices, numbers or names not found in the source, long copied runs, or a missing business name.
4. We rehearsed each batch on a copy of the live data, then applied it, keeping a record so it could be rolled back.

By the end of that pass, 6,165 listings had a profile written from the business's own site. Contact details, where the business published them and our record was empty, were filled in for 4,739 of them.

Not every listing could be written. The owner's instruction was that listings with no website, or with problems, should be taken offline, reversibly, unless we bring them back later. We parked 15,238. The main reasons:

• no website on the listing: 9,392;
• a website that no longer answers: 3,112;
• a domain now owned by someone else, parked or for sale: 667;
• a site that was not this business, unreadable, too thin, off-topic, out of scope or behind a bot wall: 2,037;
• a site saying the business had closed: 33.

We did not park IT service providers during the experiment, claimed listings, or listings with reviews. We also restored 104 that had been parked by mistake: they were rebrands, moved to a new domain, whose first draft had failed our checker.

Parked listings return a not-found page. Archiving is reversible, and every batch has a record we can roll back.

Checking the rest of the catalogue's links

Next we checked the website of every live, non-thin listing, 77,998 of them. On 27 September about 57,200 answered normally, roughly 2,700 had no domain record, about 630 were parked, roughly 3,900 returned errors and roughly 4,650 had moved.

We removed 2,675 listings whose domains had no record or were parked, excluding IT service providers, claimed listings and anything we had rewritten. Another 4,530 stayed online but flagged, so their pages never link to a dead site. The contact card shows the dead address struck through, with the date we checked it.

Round 2: every live listing to one standard

On 28 September the goal widened: rewrite the historical and recent listings to a single standard, and be hard on the thin pages Google lands on. The AdSense review is the bar, and we treat the rehearsal as worthwhile for our own improvement whatever the outcome.

The standard is deliberately plain. A short tagline. A body of 150 to 300 words, or 110 to 150 where sources are thin. An opening paragraph naming the business, a section on what it offers with three to six bullets, an optional section on who it works with, and a key-facts section holding only stated facts. Third person, British English, no hype, no money amounts, no contact details, and nothing copied. Writers can decline a listing for one of four reasons: unrelated, not enough to go on, out of scope, or closed.

On 29 September, in our production records:

• 29,722 of 72,564 live, active companies had been rewritten in our own words;
• only 238 of 8,561 products, because the pool was mostly companies;
• writers had skipped 2,325 listings, as the brief allows, so those keep their old text for now;
• 89 listings were still rejected after two fixes and are with us to rewrite by hand;
• roughly 19,000 more listings were still waiting in the publishing queue.

What this means for people using TrustList

If you are a buyer, more of what you read was written from a business's own current site by a process with rules, and less is copied from a page that may have changed hands. Fewer pages exist, and we think that is the better trade.

If you run a business listed here, it may have been rebuilt, flagged or parked. If your profile is wrong or missing, you can contribute a correction, and our vendor page explains claiming. We do not sell rankings.

If you want to understand how we decide what to trust, our trust page is the place to start, and the rankings index shows where the results of all this land.

Other articles in this series look at particular parts of the story: the clean-up of spam and junk, how we write listings, our rules for guest posts and listings, what we have learned about insights and AI-assisted drafting, and why we keep our language plain. Read them alongside this one.

Before and after, in numbers

Here is where things stood when the AdSense refusal arrived, and where they stood a few days later. Every figure carries the date of our record.

• Thin profiles: around 17 September 2026, 24,350 live listings (about a quarter) had fewer than 50 words of profile text. By 29 September, 29,722 of 72,564 live company listings had been rewritten to one standard, with about 19,000 more queued.
• Voice: around 17 September, 82% of the older company profiles with 50 or more words (54,541 of 66,473) were written in the business's own voice ("we", "our"). On 28 September, about 11,500 of 81,127 live, active listings were in our own words (14%). By 29 September the rewritten count above had passed 29,000.
• Spam: before 26 September, 1,719 listings were flagged by our scan. After review, 164 wholly spam and 24 partly spam listings and 52 paid guest posts were archived on 26 and 27 September, and 7 reviews were blocked.
• Thin listings with nothing to source: 15,238 were parked on 27 September (archived, and reversible), while 6,165 were rewritten from the business's own site.
• Dead links: 2,675 listings whose domain had no DNS record or was parked were taken offline on 27 September, and 4,530 more were flagged so their pages never link a dead site.
• Rankings: on 25 September about 18,000 ranking pages sat in the sitemap, many built from import tags. On 28 September a /best page existed only where a published ranking backed it, about 12,000 of them; category pages without one answer 404 and nothing links to them.
• Category pages: 68,678 category-and-term pairs had a page, but only 11,962 had five or more listings that were not thin.

The change is one of rule as much as size. Before, a page existed because an import had created it. Now a page stays because something real stands behind it: a profile written to a standard, or a ranking with enough substantial listings to justify it. Each rewritten profile covers what the business offers, who it works with and its key facts, uses only facts from the business's own site and our records, and passes automated checks for length, copying, hype, invented numbers or names, prices and contact details.

For a reader, that means fewer pages and more that can be relied on. A profile is less likely to be a business's marketing copy or the text of a domain that changed hands. A ranking is less likely to be padded with empty entries. Some of the numbers above will have moved by the time you read this, which is why we date each one.

Sources

• TrustList: listing guidelines, contribute, trust, for vendors, rankings and our earlier article on search traffic.
• Google: AdSense programme policies, spam policies for Google web search and creating helpful, reliable, people-first content.
• Figures about TrustList come from our own internal records, dated in the text: the AdSense response and sitemap counts of 25 September 2026, the spam scan of 26 September, the fast pass and link check of 26 and 27 September, and the round 2 counts of 29 September 2026. They are snapshots and will have moved on.