The best way to test a blog content service is a small paid pilot: one real topic, one written brief, one deadline and a scorecard you fill in before you see the invoice. A portfolio shows what a provider can do on its best day with a topic it chose. A pilot shows how it works with your business, your facts and your feedback.
This matters because blog writing is easy to sell and hard to compare. Two providers can quote the same price per post and deliver very different things: one gives you 1,200 words of confident generalities, the other a researched piece that a customer, and an AI assistant, could quote. Below is the five-step trial we would run if we were buying, with a scorecard you can copy.
Step 1: What should you ask for before the pilot?
Before you send any brief, get clear answers to five questions. A provider that dodges them has told you something already.
- Who writes the first draft: a human, an AI tool or a combination?
- Who edits it, and is that a different person from the writer?
- How are facts and numbers checked, and are sources listed?
- Who owns the finished text and any research files?
- What happens when you cancel: do posts stay online, and is anything held back?
There is no single right answer to the first question. Plenty of strong providers use AI for drafting and humans for judgement. What matters is that they say so plainly. For the commercial side of what a retainer should include, see our guide to what a blog content retainer should include.
Step 2: How do you design a fair pilot?
Use the same conditions for every candidate:
- A real topic. Choose a question your customers ask, ideally one where you know the correct answer well enough to spot errors.
- A written brief. One page: audience, goal, the main point you want made, two or three facts that must appear, and what to avoid.
- A fixed deadline. Five to seven working days is normal for a 1,200 to 1,500 word post.
- A small fee. Paying for the pilot signals that you are serious and gets you honest work instead of a recycled sample. If a provider refuses to do a paid pilot at all, ask how they would approach month one.
Keep the brief identical between providers. If you give one of them a call and the others only an email, you are testing your own generosity, not their quality.
Step 3: What should you look for in the draft?
Print it, or read it slowly on a screen, and score each point from 1 to 5.
| Criterion | What a 5 looks like | What a 1 looks like |
|---|---|---|
| Accuracy | Every claim about your business and industry is correct or clearly sourced | Confident errors or invented figures |
| Specificity | Concrete examples, numbers framed honestly, named steps | “In today’s fast-moving digital landscape” filler |
| Direct answer | The core question is answered within the first two paragraphs | Three paragraphs of preamble |
| Structure | Question-style headings, short paragraphs, lists only where they help | A wall of text or bullet-point overload |
| Voice | Sounds like a knowledgeable person in your field | Sounds like every other blog |
| Edit effort | You change a few lines | You rewrite whole sections |
The last row is the most honest one. A draft that costs you two hours of fixing is not cheap, whatever it cost to buy. Our editing checklist for AI-written drafts lists the issues that most often need correcting, and it works just as well on human drafts that came out generic.
Step 4: How do you test the process, not just the text?
The draft is only half of what you are buying. Use the pilot to watch four behaviours:
- Questions. A good provider asks two or three sharp questions before writing. One that asks nothing is guessing.
- Handling corrections. Plant one deliberate factual correction in your feedback. See whether it is fixed everywhere it occurs and whether they explain what went wrong.
- Source honesty. Ask where a particular statistic or claim came from. A good answer is a link or “that is our estimate, and here is how we reasoned it”. A bad answer is silence or a made-up reference.
- Turnaround and communication. Note whether deadlines are met and whether problems are flagged early.
These behaviours predict month six far better than the first draft does.
Step 5: How do you check the post works for search and AI answers?
A blog post that reads well can still be hard for search engines and AI assistants to use. Check a few structural points:
- Does the opening paragraph state the answer in a sentence you could quote?
- Are the headings phrased as the questions people ask?
- Is each claim specific enough to be useful on its own, rather than vague?
- Are there links to related pages on your site, with descriptive anchor text?
- Does the post avoid duplicate or thin sections added to hit a word count?
If you are weighing a content service against doing it yourself, our comparison of DIY ChatGPT content and a content service sets out the cost side of that choice.
What are the warning signs that should end the conversation?
- Promised rankings or traffic numbers within a fixed number of weeks. Nobody controls that.
- Guaranteed “AI citations”. Nobody controls that either.
- No named human responsible for your account.
- Posts that could be published on any company’s site with the name swapped.
- A contract that locks you in for a year with no exit clause and no review point.
- Reluctance to show how they handle sources or editing.
What should the pilot cost, and how long should it run?
Keep it proportionate. One post, or at most two, is enough to see the pattern. Expect to pay the provider’s normal per-post price, or a modest discount; be wary of providers who want to turn a pilot into a three-month commitment before showing anything. If you are comparing three candidates, budget for three pilot posts and treat the spend as the cheapest insurance in the whole project. A bad six-month contract costs many times more, and the posts you fund in the pilot are yours to use if they pass your own edit.
It also helps to run the candidates in parallel, so you judge them in the same week and under the same pressure from your own calendar. Sequential trials tend to favour whichever provider came last, because by then your brief has improved.
How do you decide after the pilot?
Add up the scores, then add three non-scored questions: Would I trust this person with a customer-facing message? Did I learn something from how they worked? Would I be comfortable if my customer read this under my name? If the answer to the last question is not a quick yes, the provider is not ready to speak for you.
Finally, set a review point in whatever you sign. Three months is a fair period for a blog programme to show quality and consistency, even though ranking effects take longer. Agree in advance what you will review: published output, accuracy, edits needed, and whether the topics are matching real customer questions. For the broader decision about whether blogging suits your business at all, see is blogging still worth it in 2026.
The bottom line
Do not buy a blog programme on the strength of a pitch deck. Give two or three providers the same brief, pay for a small pilot, score the result with the table above, and watch how they behave when you push back. A week of testing costs very little next to six months of generic content.