Filling a dead blog: six runs, measured
This is a measurement report, not a case study. Every figure below came out of a real run on durbun.net between 16 and 21 September 2026. There are no traffic numbers here, because a blog that was empty last week does not have any yet. What it does have is a cost, a duration and a quality score per article, and those are the numbers that decide whether leaving a campaign running is sensible.
The starting point
DURBUN's own blog held exactly one published post. That is the situation the Auto Publisher is built for: a site that is alive, has a topic, and has nobody with an afternoon spare to write for it. Rather than describe what a campaign would do, the six runs below were actually executed against the live service, on the live model providers, paying real credits.
What six runs cost
| Article | Template | Words | Images | Credits | Time |
|---|---|---|---|---|---|
| Best AI Auto-Blogging Content Generator | SEO-friendly | 1,421 | 0 | 1,521 | not timed |
| How to Choose an AI Article Generator for WordPress | SEO-friendly, short | 792 | 3 | 3,792 | 90 s |
| Is DURBUN a Good AI Article Generator for SEO Content? | SEO-friendly, short | 549 | 3 | 3,549 | 87 s |
| AI article generator pricing per 1,000 words | Programmatic SEO | 792 | 0 | 792 | 96 s |
| AI Article Generators That Publish to WordPress Automatically | SEO-friendly | 1,311 | 4 | 5,311 | 132 s |
| How to Run an Auto Blogging Campaign Without Thin Content | SEO-friendly | 1,358 | 4 | 5,358 | 165 s |
Across the five runs that were timed: 4,802 words, 18,802 credits and 9 minutes 30 seconds of machine time in total. No run failed, and no run needed a second attempt.
Where the credits actually go
One credit is one output word. That part is small. The images are what move the number: at the low quality setting an image is 1,000 credits, and a full-length article carries three to six of them. The pattern in the table is consistent.
| A 1,300 word article | Credits |
|---|---|
| Words only, images pulled from Pexels or Google | 1,300 |
| Plus four AI images, low quality | 5,300 |
| Plus four AI images, medium quality | 9,300 |
| Plus a YouTube embed, when one is found | add 100 |
So the first decision that matters is not the article length. It is whether the images are generated or fetched. Fetched images cost nothing and the article still arrives illustrated.
What a schedule costs
Taking the 1,300 word shape from the table and multiplying it out:
| Cadence | Fetched images | Four AI images, low |
|---|---|---|
| One a day | 39,500 credits a month | 161,200 credits a month |
| Three a day | 118,600 a month | 483,500 a month |
Credits are bought once and never expire, so these are running costs rather than a subscription. The rate per 1,000 credits is on the pricing page and changes rarely.
Machine time, and what it means for a schedule
The timed runs took between 87 and 165 seconds. That is the whole pipeline: web research, outline, drafting, two rewriting passes, image generation, image placement and saving. Three articles a day is therefore about eight minutes of work a day, none of it yours, and the schedule never collides with itself because a campaign runs one article per slot.
What the output actually scored
The published article was then put through the Article Quality Report, which grades a draft the way a search quality rater would, here on two models at once. It scored 6 out of 10. That is worth reading carefully rather than celebrating or panicking about: the grader reserves 9 and above for pages it would rank above the current results, and it puts most published writing on the web at six or seven. A six is a normal, publishable article.
It was strongest on search intent (7) and weakest on evidence and credibility (4.5), and it named three specific things to fix: a comparison that lists tools without separating them, a price quoted for one product and not the others, and an FAQ repeating what the body already said. Those are editing notes, and they took twelve seconds and 152 credits to produce.
What this report does not tell you
It does not tell you what the articles will rank for, how much traffic they will bring, or whether they will be indexed. Six articles over six days is not a sample anyone should draw a traffic conclusion from, and any figure offered here would be invented. Come back when there is a quarter of data behind it.
What it does tell you is the part that is knowable today: a full article, researched, illustrated and formatted, costs between 800 and 5,400 credits and lands in about two minutes, consistently, without supervision. Whether that is worth running is a decision you can now make with real numbers instead of a promise.
Method
Every run used the live service on the live model providers. Credits are read from the account balance before and after each run, not estimated. Durations are wall clock from the moment Generate is pressed to the moment the article is saved. The quality report is the unedited output of the Article Quality Report on the published article. Word counts are the generator's own and run a little under a rendered count, which is why the quality report gives a higher figure for the same article. Nothing here was rounded in DURBUN's favour.