AI content strategy: what it is and where AI fits
Short answer An AI content strategy is a plan that says which content jobs AI does and which a person does. AI is good at research, outlines, first drafts and ad variations. People supply experience, prices, local facts and the final check, which Google asks for. Personalizing needs customer data and consent, so most small businesses start with simple segments.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callWhat is an AI content strategy?
It is a short written plan for your website, emails and ads that answers three questions: which jobs AI does, which jobs a person does, and who checks the result before it goes out.
The tool is not the strategy. A plumber with ChatGPT and no plan publishes generic posts. The same plumber with a plan uses AI for the slow parts and spends their own time on what only they know.
| Stage | AI does well | You do | The rule that applies |
|---|---|---|---|
| Plan | Research a topic, group customer questions, build an outline | Pick topics from real customer questions | None; this is the safest use |
| Write | First drafts, titles, meta descriptions, rewrites | Add prices, local facts, your experience; fact-check | Google: review all AI content before publishing |
| Personalize | Ad variations, versions of an email for each customer group | Choose the groups; get consent | CAN-SPAM (US), CASL (Canada) |
| Measure | Summarize reports, spot pages losing visits | Decide what to fix | None |
Where AI fits in planning
Planning is where AI saves the most time with the least risk. Google says so directly:
"Generative AI can be particularly useful when researching a topic, and to add structure to original content." (Google Search Central: Guidance on using generative AI content)
The catch: ask AI "what should a dentist blog about?" and you get the same list every dentist gets. Feed it your own material instead: the questions callers asked last month, emails from customers, notes from estimates. Ask it to group them and turn each group into an outline. Your topics then come from your customers, not from the tool.
Whether you need a blog at all is a separate question, covered in do blogs help SEO?.
Where AI fits in writing
AI can write a first draft in minutes. It cannot know your prices, your service area, what went wrong on last week's job, or a rule that changed after it was trained. That part is your job, and Google expects a person to do it:
"It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." (Google Search Central: Guidance on using generative AI content)
To keep drafts sounding like you, give the tool a few of your past emails or pages as examples, then edit in your own words.
The same Google page says the review also covers page titles, meta descriptions (the short text under your link in Google), structured data and image alt text, not just the article.
Google doesn't penalize a page for being AI-written; it judges quality (what Google's AI policy says). On telling readers, Google's helpful-content guide suggests a note when people would wonder:
"Overall, AI or automation disclosures are useful for content where someone might think "How was this created?" Consider adding these when it would be reasonably expected." (Google Search Central: Creating helpful, reliable, people-first content)
Where AI fits in personalizing
Personalizing means showing different groups of customers different messages. The big promise, AI predicting what each person wants, needs far more data than most small businesses have. Google Analytics 4's predictive audiences, for example, only work when, within the last 28 days, at least 1,000 returning users did the action you're predicting and at least 1,000 didn't (Google Analytics Help: Predictive metrics). A local service business rarely gets there.
What works at small-business size:
- Groups you already know. New leads, past customers, people who got an estimate but didn't book, or customers by service type. AI writes one version of an email for each group; you check it. More in email and text marketing.
- Let the ad platform do it. A Google responsive search ad takes up to 15 headlines and 4 descriptions, then tests combinations and learns which are most relevant for different searches (Google Ads Help: About responsive search ads). AI can draft the headlines; Google does the matching.
Using customer data brings rules:
- US: CAN-SPAM does not require consent before a marketing email, but every one must tell people how to opt out, and you must honor an opt-out within 10 business days (FTC: CAN-SPAM compliance guide).
- Canada: CASL bans sending a commercial email or text unless the person "has consented to receiving it, whether the consent is express or implied" (Canada's Anti-Spam Legislation, section 6). See CASL consent.
What not to hand to AI
- Reviews and testimonials. The FTC's 2024 rule bans fake reviews, including ones posing as someone who doesn't exist, such as AI-generated reviews (FTC, August 14, 2024). See asking for reviews: the rules.
- Mass pages. Fifty near-identical city pages written by AI fit Google's definition of spam: "Scaled content abuse is when many pages are generated for the primary purpose of manipulating search rankings and not helping users." (Google Search Central: Spam policies). See city pages for service areas.
- Facts you haven't checked. Prices, laws, licensing, warranty terms, medical or legal claims.
- Your judgment. What you recommend and why. That experience is what makes a page worth reading (E-E-A-T).
How to write your own AI content plan
- List the content you already make or want: service pages, a monthly email, ads, review replies.
- For each, mark the AI jobs (research, outline, draft, variations) and the human jobs (facts, examples, final check).
- Name the checker. One person signs off on everything before it goes live.
- Write down your groups for personalizing and confirm you have consent to email or text each one.
- Once a month, look at what brought calls or bookings, and make more of that.
Example
A roofing company, one page a month
The owner records answers to five questions customers ask about storm damage. AI turns the recording into an outline and a draft. The owner adds two real job photos, local permit costs and their warranty terms, then checks every fact. About 2 hours a month instead of 8 for writing from scratch.
A dental office, three email groups
The office sends one checkup reminder a month. AI drafts three versions: patients due for a cleaning, patients with unfinished treatment, and new patients from the last 90 days. The office manager checks each. Every email carries an unsubscribe link; for patients in Canada, the office confirms consent first.
FAQ
Do I need special AI content software?
No. A general chatbot handles research, outlines and drafts. Paid tools mostly add publishing and scheduling. The plan and the human check matter more than the tool.
Will Google rank AI-written pages?
They can, if they are useful and accurate. Google judges the content, not how it was made, but treats mass low-value pages as spam.
Should I tell readers I used AI?
Google suggests a disclosure where readers would reasonably wonder how a page was made. For a page you researched, rewrote and checked yourself, it is your call.
Can AI personalize my website for each visitor?
Some tools can, but they need steady traffic and data to learn from. Most small businesses get more from separate pages for each service and emails written for each customer group.
Different for you?
- You want AI to run your SEO end to end: can AI do SEO?
- Old pages losing visits: update, merge or delete old pages
- Email or texts to customers: email and text marketing
- You'd rather have the content work done for you: SEO