TL;DR
Thorsten Meyer AI reported that a 28-day audit of its 474-site WordPress publishing network found 80% of posts going to 38 sites, while 249 sites received no posts. The company attributed the imbalance to separate supply and placement problems across Stenvrik and DojoClawAI, and said it has added caps, recency ordering, feed changes and higher fan-out limits.
Thorsten Meyer AI said a 28-day audit of its 474-site WordPress publishing network found that 80% of posts were landing on just 38 sites while 249 sites received none, exposing a placement imbalance that total output metrics had failed to show.
The audit covered a network fed by two systems: Stenvrik, described by Thorsten Meyer AI as a news-intelligence layer that ingests feeds, scores stories and adds geographic tags; and DojoClawAI, described as the content engine that rewrites stories in each site’s voice and distributes them across the catalog.
According to the engineering note, the top 38 sites represented about 8% of the catalog but carried 80% of output. The four busiest sites, all technology titles, each received more than 200 articles per week. At the other end of the catalog, 249 sites, or 53% of the network, had zero posts during the audit window.
Thorsten Meyer AI said the failure was not caused by a single bug. It attributed the pattern to two separate causes: DojoClawAI was repeatedly selecting broad technology sites once stories cleared the relevance gate, while Stenvrik was supplying far more technology and AI material than the catalog could absorb. The note said 53% of supplied content was tech or AI, while only about 13% of sites were in those categories.
Why It Matters
The report matters because it shows how automated publishing systems can appear healthy when measured only by total production. In this case, throughput remained active, but distribution across the network became highly uneven. For operators of large site networks, that gap can affect freshness, audience coverage, category balance and the usefulness of automation metrics.
The case also points to a broader risk in automated content operations: relevance alone may not produce balanced output. Thorsten Meyer AI said every individual placement could be considered correct, yet the aggregate pattern left half the catalog inactive. That distinction is material for publishers relying on automation to manage large inventories of sites, topics or local editions.
WordPress for Beginner: A Step-by-Step Guide to Creating and Managing Modern Websites Using Themes, Plugins, and AI Tools No Coding Required
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
The engineering note describes the network as a 474-site catalog connected to Stenvrik and DojoClawAI. Stenvrik supplies and ranks stories; DojoClawAI handles rewriting and placement. Thorsten Meyer AI said the separation helped diagnose the issue because the imbalance came from both supply and placement rather than one shared component.
The fix described in the note has three parts. DojoClawAI now uses per-site weekly caps, a global least-recently-used ordering method and a starvation floor designed to keep idle eligible sites in the selection path. Stenvrik’s feed pool was audited for live output, broken RSS feeds were removed, and verified feeds were added across categories including home, garden, health, food, fashion, auto, science and pets. Scheduler settings were also raised, including fan-out width from five sites to seven and quota depth from two to three.
“A 474-site network quietly collapsed onto 38 of its own favorites while half the catalog went dark.”
— Thorsten Meyer AI engineering note
“The tempting move is to blame the matcher and move on.”
— Thorsten Meyer AI engineering note
“The proof is in the next weeks of data.”
— Thorsten Meyer AI engineering note
Generative AI for Software Testing: Improve QA with AI-Powered Automation
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
The results of the fix are not yet fully known. Thorsten Meyer AI said the changes affect future placement and do not retroactively change the 28-day audit period. The note lists dormant sites as expected to shrink, but it does not provide post-fix totals showing how many inactive sites have resumed publishing.
The source material is a first-party engineering note. It provides internal figures and system descriptions, but no independent audit data is included. It is also unclear how the changes will affect reader engagement, search performance, editorial quality or category-level traffic over time.
Extracting Intelligence from RSS News Feeds Using Python and AI: From Global Headlines to Actionable Intelligence
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
The next milestone is measurement over the following weeks. Thorsten Meyer AI said instrumentation is the real deliverable, with the network expected to show whether caps, global recency ordering, feed rebalancing and higher scheduler limits reduce concentration without starving relevant topics.
![Express Schedule Free Employee Scheduling Software [PC/Mac Download]](https://m.media-amazon.com/images/I/41yvuCFIVfS._SL500_.jpg)
Express Schedule Free Employee Scheduling Software [PC/Mac Download]
Simple shift planning via an easy drag & drop interface
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What was the main finding of the audit?
Thorsten Meyer AI said 80% of publishing output went to 38 of 474 sites during a 28-day period, while 249 sites received no posts.
Was this caused by one software bug?
The company said no. Its engineering note attributes the imbalance to two causes: placement concentration inside DojoClawAI and a content supply mix from Stenvrik that leaned heavily toward tech and AI.
What changed after the audit?
Thorsten Meyer AI said it added per-site caps, global recency ordering, a starvation floor, verified feeds in under-served categories and higher scheduler limits for fan-out and quotas.
Has the fix been proven yet?
No final post-fix outcome is included in the source material. The company said the changes must be judged by the next weeks of publishing data.
Source: Thorsten Meyer AI
