Online content generation is the practice of using AI tools to draft the pages, posts, emails and replies a business publishes. This guide is for owners and office managers of local service businesses in the Twin Cities East Metro, the St. Croix Valley and western Wisconsin who want the speed of a drafting tool without publishing a wrong fact under the company name.
It lays out a six-step workflow from brief to publication, gives example prompts, and covers what Google's pages say about generated content, what to keep out of a prompt, and the federal rule on fake reviews. It is general information and not legal advice, so confirm anything legal with your attorney. For the wider question of which jobs suit AI at all, read our post on practical AI for a local service business.
Online content generation with AI: a six-step workflow
The workflow keeps the tool in the middle of the process, where drafting happens, and puts a person at both ends. The table shows who does what, and the sections after it explain each step.
Scroll to see all columns
| Step | Who leads | What the AI tool does | What a person checks |
|---|---|---|---|
| 1. Brief | Owner or office lead | Suggests questions customers ask | Picks the topic, the reader and the facts to include |
| 2. Draft | The tool | Writes a first version from your notes | Confirms it used only the notes |
| 3. Fact check | Someone who does the work | Lists every claim as a checklist | Verifies each claim against a source or a job record |
| 4. Local detail | Owner or crew lead | Marks where local detail is missing | Adds towns, seasons, job details and photos from your own work |
| 5. Human edit | An editor with a name | Suggests tighter wording | Cuts filler, restores your voice, approves the page |
| 6. Publish | Whoever owns the site | Drafts the title, summary and image text | Reviews those fields and decides on a disclosure line |
Step 1: write the brief yourself
Start with who the page is for, what question it answers and which facts belong in it. A brief can be five lines in a note. Google's page on creating helpful, reliable, people-first content frames content in three questions it calls Who, How and Why. It calls the Why perhaps the most important question, and it asks whether you are making the page primarily to help people. A brief that starts from a real customer question passes that test more easily than one that starts from a keyword.
Step 2: draft from your notes
Give the tool your notes and tell it to stay inside them. These example prompts are starting points to adapt, and each one assumes you replace the bracketed text:
- Brief helper: "List ten questions a homeowner in [city] asks before booking [service]. Do not state prices, license rules, warranty terms or code requirements. Mark anything you are unsure of with [CHECK]."
- Draft: "Write a 300-word draft for a [service] page using only the notes below. Where a fact is missing from the notes, write [ADD FROM OWNER] and do not guess."
- Claims list: "List every factual claim in this draft, one per line, so a person can verify each one."
- Gap finder: "Show me where this draft is generic and a local detail from [town] or a real job would make it specific. Do not invent the detail."
The placeholders matter because of how the tools work. Google's guidance on generative AI content says generative models predict likely word sequences from training data and do not retrieve facts, which is why outputs may contain inaccuracies it also calls hallucinations. A tool that is told to flag its gaps gives you less to hunt for later.
Step 3: check every fact
Google's guidance calls it critical to review and fact-check generated content before publishing, and says the same review applies to title elements, meta descriptions, structured data and image alt text. At this step, a person who does the work should read the claims list and tick each one against a source: a manufacturer page, a government page, or your own job records. Anything with no source comes out. Prices, warranty terms, license statements and code references get the closest look, because a wrong one reads as a promise from your company.
Step 4: add the local detail only you have
This is the step that turns a generic draft into a page worth reading. A tool has no way of knowing which streets flood after a spring thaw in your part of the East Metro, how your crew schedules around the first hard freeze, or what you found behind the wall on a job last month. You do. Add job details, real photos from your own work and the towns you actually serve. Our Brand & Copy service covers site copy, blogs and proposals.
Step 5: edit it as a person with a name
Read the page aloud, cut filler, and put your own words back where the draft turned generic. Give the page a byline that leads to a real person. Google's page says it helps people understand the E-E-A-T of content when it is clear who created it, and it strongly encourages accurate bylines where readers might expect them. It also warns against fabricated creator profiles, such as made-up names or AI-generated headshots, and says any form of deception is a signal of a low-quality page.
Step 6: publish, and say how it was made where readers would expect it
Review the generated extras too: the title, the meta description, the image text. Then decide on disclosure. The same Google page asks, for content where automation is used to substantially generate it, whether the use of automation is self-evident to visitors through disclosures or in other ways. It says AI or automation disclosures are useful where someone might think "How was this created?" and suggests adding them when it would be reasonably expected. For a service business, a short line near the byline is one reasonable choice, and that is our suggestion. Google's page does not prescribe a form.
What Google says about generated content at volume
Google's spam policy does not bar generative tools as such. Google's spam policies define scaled content abuse as many pages generated mainly to manipulate search rankings and not to help users, and they say that applies however the pages are made. Using generative tools to produce many pages without adding value for users is one of the examples the policy lists. The generative AI guidance repeats that point, so the line is about purpose and value.
For a local business, the practical version is a question of scale. In our reading, ten accurate pages about services you perform and areas you serve are not what the policy describes, and two hundred near-identical town pages produced from one prompt resemble the examples it lists. If you are weighing an agency that offers volume, our guide to judging a content generation agency covers the questions to ask.
What AI tools get wrong about Minnesota and Wisconsin
A tool writes fluent text about a rule without telling you which state the rule belongs to or whether it applies to you. These are checks a person in the East Metro should make before a page goes live, and the list is partial.
- License numbers. Minn. Stat. 326B.87 says a licensee's license number must appear in advertising, including websites. The Department of Labor and Industry's residential contractor FAQ says the same for websites. Under 326B.802, a licensee for those sections is a residential building contractor, residential remodeler, manufactured home installer or residential roofer. A draft may leave the number out or state the rule as if it covered every trade, so check who the rule covers before you copy it onto a page.
- The Wisconsin side. Wisconsin's Department of Safety and Professional Services keeps its own A to Z professions list, which names credentials such as Dwelling Contractor, HVAC Contractor and Master Plumber. A business that serves both Hudson and Woodbury should be clear about which state's credential applies to which work.
- Service areas. Google's service area help page lets a business list up to 20 areas by city, postal code or another area, says a radius cannot be used, and says the overall area should stay within about two hours of driving from the base. A draft that promises a 50-mile radius will not match the service area you can set on a Google profile, which cannot use a radius and should stay within about two hours of driving from your base.
- Seasons and timing. A tool writes for an average climate. Minnesota's freeze and thaw timing, your own scheduling and your lead times are facts only your crew can supply.
- Warranties, prices and code. Keep these out of drafts unless a person has typed them in from your own documents.
This is general information and not legal advice. For a license or advertising rule, confirm with the state agency or your attorney.
Keep customer details out of the wrong plan
Prompts are easy to fill with real names, addresses and job notes, and the plan you use decides what can happen to them. OpenAI's help page on model improvement, as it read in November 2026, says its individual plans (Free, Plus and Pro) may use your content to train its models, with a setting to opt out. It says feedback on a response, such as a thumbs up or down, may let the whole conversation be used even after you opt out. It also says ChatGPT Business, Enterprise, Edu and its API do not use inputs or outputs to improve models by default.
That page covers OpenAI's plans only, and other vendors have their own terms, so read the one for the tool you use. Our suggestion is to write prompts with placeholders such as "[customer]" and "[address]" and keep the real details in your CRM. A draft rarely needs a customer's name to be useful.
Never use a tool to write a review that pretends to come from a customer
The Federal Trade Commission's rule on consumer reviews and testimonials is 16 CFR part 465, published in the Federal Register on August 22, 2024. Section 465.2(a) makes it an unfair or deceptive practice for a business to write, create or sell a consumer review or consumer testimonial that materially misrepresents that the reviewer exists, that the reviewer used the product or service, or what the reviewer's experience was. The section covers consumer testimonials as well as reviews, and the definitions describe a testimonial as a promotional message that consumers are likely to believe reflects a real consumer's experience. A customer quote written for your own website is worth running past your attorney.
The practical line for AI is simple. A tool can help you draft a polite reply to a real review or tidy a question you ask customers. It should never produce a review or a quote that pretends to come from a customer who does not exist. Ask your attorney if you are unsure how the rule applies to a particular practice.
If you want site copy, blogs or proposals written to sound like your business, contact McKinney Creative Ventures.