AI for content creation that still sounds like you

AI for content creation works when the software writes from material only your business has: the replies your team sends customers, the notes from real jobs, the questions people ask on sales calls. Given that, it can draft articles, newsletters and social posts, turn one good piece into several, keep a content calendar moving and refresh older posts, while a person reads and approves everything before it goes out. Given nothing but a topic, it produces the flat, interchangeable text people now call AI slop.

That difference decides whether AI content creation helps your business or dilutes it, and the rest of this post is about getting the first kind.

At Sanaf this is the AI content assistant job, one of the jobs on the roster a business can give Sanaf. Where the job description says something plainly, this post quotes it.

Why so much AI content sounds the same

Ask a language model to "write a blog post about choosing a roofer" and it gives you the most likely blog post about choosing a roofer. It has read thousands of them. The result is average by construction: accurate enough, organised, polite, and indistinguishable from every other roofer's site that asked the same question.

The usual fix is to tell the tool to "sound more human", and a whole category of humanize AI text tools exists to do that after the fact. They swap words and vary sentence length. What they cannot add is the thing that was missing to begin with, which is something only you know.

Compare two openings for the same post.

The generic one: "Choosing the right roofing contractor is one of the most important decisions a homeowner can make."

The one written from your sent folder: "Most people call us after the second leak, not the first, because the first one stopped when the rain did."

The second sentence came out of a reply somebody on your team wrote to a customer. No model would have produced it from a topic, because it is a fact about your customers. That is what brand voice mostly is in practice. Tone guidelines help, and a short list of words you never use helps more, but the voice comes from specific things said by people who do the work.

The job page makes the same point about where content comes from: "The clearest explanation of what you do is usually a reply somebody wrote to a customer at speed, and it is read by one person and then buried."

Does Google penalize AI content?

This is the question most owners ask before anything else, so it is worth answering from the source.

Google published its position in February 2023, in a Search Central post titled "Google Search's guidance about AI-generated content". The short answer is no, not for being AI-generated. In Google's words: "Appropriate use of AI or automation is not against our guidelines." The same post says its ranking systems "aim to reward original, high-quality content that demonstrates qualities of what we call E-E-A-T: expertise, experience, authoritativeness, and trustworthiness," and the section heading is "Rewarding high-quality content, however it is produced."

What Google does act on is purpose and scale. The same post: "Using automation, including AI, to generate content with the primary purpose of manipulating ranking in search results is a violation of our spam policies." Its current spam policies name the pattern as scaled content abuse, "when many pages are generated for the primary purpose of manipulating search rankings and not helping users," and give "using generative AI tools or other similar tools to generate many pages without adding value for users" as an example. (Both pages checked on 29 September 2026.)

So the risk with AI generated content is publishing a hundred pages that say nothing new. Content built from your own experience is the kind the E-E-A-T framework asks for, because the first E is experience, and your team has it and a model does not.

That also answers the related question of whether AI content is bad for SEO. Thin content is bad for SEO. AI makes thin content cheap to produce, which is why so much of it exists.

What AI content creation actually covers

"Content" covers several different jobs, and AI helps with each in a different way. The AI content assistant job lists six duties:

• Creates blog posts
• Creates website content
• Creates newsletters
• Repurposes existing content
• Manages content distribution
• Maintains content calendars

Social posts sit with the neighbouring AI social media manager job, which builds the social calendar, writes posts, repurposes content and answers comments. Search research and content briefs sit with the AI SEO specialist. Sanaf can hold all three jobs at once, so in practice they share the same source material and the same calendar.

Blog posts and website pages

This is where most businesses start and where most stop. A blog post takes an afternoon that nobody has, so the blog gets three posts in a burst and then goes quiet. The job page names what that looks like from outside: "A dated archive tells a visitor something about the business that nobody intended to say, and it tells a search engine something similar."

An AI draft removes the blank page. The person who knows the subject reads a finished draft, corrects what is wrong, adds the story only they know, and approves it. That is a twenty-minute job instead of an afternoon.

Newsletters

An AI newsletter works the same way. The month's material already exists: jobs finished, questions answered, a price change, a new service. The draft pulls it together from your records and the email list you already have in Mailchimp, and a person sends it.

Content repurposing

Content repurposing means turning one piece into several shapes for several places. A long answer to a customer becomes an article. The article becomes a newsletter section. Three of its points become three social posts. A finished job with good photos becomes a short case write-up and a week of posts.

This is the part of content production that almost never happens by hand, because it is tedious and the original already feels "done". The job's third signal describes it exactly: "Good content gets written once and never reused." Repurposing is also where AI is most reliable, since the facts and voice are already in the source and the work is mostly reshaping.

Refreshing older posts

A post that was right in 2024 can be wrong now. Prices changed, a product was renamed, a better example came along. The job page is blunt about the value: updating dated pieces "is usually better value than another new article, and it is almost always the thing that gets skipped because nobody feels the absence of it."

Evergreen content, the posts that answer a question people will still be asking in three years, only stays evergreen if somebody checks it. An AI content assistant can keep a list of what is due for review, draft the changes, and say what changed and when.

The content calendar

A content calendar is the schedule of what goes out, where and when. Most businesses have one in a spreadsheet, and most of those stopped being updated in the same month as the blog. Keeping it is the dullest part of the whole process, which is why it suits software. The calendar is also what makes publishing steady instead of a burst every few months, and steady is what both readers and search engines notice.

A content production process that keeps your voice

Here is the content creation workflow that works, in the order it happens.

1. Start from your own material. Point the job at the sources that already hold your voice: the documents in Google Drive and Google Docs, your website, your past newsletters, and the questions customers keep asking. The job's first signal is "Customers ask the same questions your website never answers." Those questions are the topic list.

2. Write down the few rules of your voice. Keep it short. The words you never use, how formal you are, whether you say "we" or the company name, how you talk about price. Brand voice guidelines that fit on one page get followed. Ten pages do not.

3. Draft from a real example every time. Each piece starts from something that happened: a real question, a real job, a real reply. That single rule does more for the voice than any tone instruction.

4. A person reads every piece. Every piece is drafted for review. This is the step people are tempted to drop once the drafts look good, and it is the one that matters most. The next section explains why.

5. Publish and distribute. Once approved, the piece goes to your site through WordPress, to your list, and to your social channels in the shapes that suit each one.

6. Revisit. The calendar includes old posts as well as new ones. Something published eighteen months ago gets checked, updated and re-dated.

Here is how that division of labour looks:

The sentences a person always checks

The job has one rule it works under, and it is short enough to quote in full: "Every piece is drafted for review. You never publish a claim about pricing, guarantees, or results without a person confirming it is true."

The second sentence matters most. The landing page explains why: "Those are the sentences that turn into commitments, and a confident paragraph written by software is exactly the wrong place for a commitment to appear by accident."

A language model writes fluently whether or not it knows something. Ask for a post about your service and it may add "most repairs finished same day" because that is what such posts often say. If it is not true of your business, you have now promised it in public. So the rule is that a person confirms any price, any guarantee and any result before it goes out. The same goes for customer names, numbers, and anything that sounds like a testimonial.

This fits the rule Sanaf works under across every job: it asks before anything it cannot undo, and it keeps a record of what it did. A published post is hard to take back once people have read it.

AI content creation tools and an AI employee

There are two ways to bring AI into content work.

The first is a set of AI content creation tools: a writing assistant, an AI content generator for social captions, a scheduling app, a separate tool to repurpose video. Each does one step well. Someone on your team still has to decide what to write, feed each tool, move the output between them, remember the calendar and chase approvals. Those tools make writing faster, but the process still depends on a person having a spare afternoon.

The second is to give the whole job to an AI employee. It keeps the calendar, reads the source material, drafts, asks for review in the channel where your team already works, repurposes the approved piece, and puts the refresh list in front of you when posts come due. You still decide what goes out under your name. You stop being the person who remembers that it should.

Sanaf does this inside Slack or Microsoft Teams. A draft arrives in a thread, somebody reads it, asks for a change or approves it, and the approval is on the record beside the draft. If you want the longer explanation of what an AI employee is, this post covers it.

An example month

Here is how it looks for a home services business with a website nobody has updated since last spring.

In the first week, Sanaf reads the last three months of customer email and pulls out the questions that came up more than a few times. Six of them have no answer on the website. It proposes a calendar: one article a week built from those six, a monthly newsletter, and two social posts a week drawn from the articles.

The first draft lands in the marketing channel on Tuesday. It opens with a line from a real reply the office manager sent in July. The owner reads it, fixes one detail about warranty terms, and approves it. The warranty sentence is exactly the kind the rule is for, so the draft had it marked for a person to confirm.

On Thursday the approved article becomes three social posts and a short section for the newsletter. Those go for review too.

In the third week, Sanaf flags two old posts from 2024 that quote prices which have since changed, drafts the updates, and adds an "updated" date to each.

At the end of the month the newsletter goes out with the month's articles, a finished job with photos, and one answer to a question customers kept asking. Every draft, edit and approval is in the record.

Frequently asked questions

How do you use AI for content creation without it sounding generic? Give it your own material to write from: real customer questions, replies your team has sent, notes from finished work. Write a one-page list of voice rules, start every piece from a real example, and have a person who knows the subject read and approve it before it goes out.

Does Google penalize AI content? Not for being AI-generated. Google's February 2023 guidance says "appropriate use of AI or automation is not against our guidelines" and that it rewards quality "however it is produced." It does act on content generated mainly to manipulate rankings, which its spam policies call scaled content abuse.

What is content repurposing? Turning one piece of content into several formats for several places, such as an article into a newsletter section and a few social posts. It is one of the most useful things AI does in content work, because the facts are already approved and the job is reshaping them.

What is a content production process? The steps a piece goes through from idea to published and maintained: choosing the topic, drafting, review, approval, publishing, distribution and later updates. AI can take the drafting, repurposing and scheduling; a person keeps review and approval.

Can an AI content assistant publish to my website on its own? Not at Sanaf. Every piece is drafted for review, and publishing happens after a person approves it. Claims about pricing, guarantees or results must be confirmed by a person first.

What does it write from? Your website, your existing documents, your email list and past newsletters, and the questions your customers ask. The more of your own material it has, the more the result sounds like you.

How much does it cost? Sanaf starts free with $120 of credit and no card. After that you buy credits as you need them and pay for the work done, not per seat. The pricing page has the details.

The short version

AI makes content cheap to produce, and that is why so much of it is worth nothing. The content that still works is written from experience, checked by someone who knows the subject, published steadily and kept current. AI can do most of the drafting, repurposing and scheduling if it writes from your material. A person still reads every piece and confirms every promise.

Read the full AI content assistant job, see every job on the roster, or start free and give Sanaf the questions your customers keep asking.