Key takeaways
- Google does not penalize content for being AI-assisted; it acts against pages made at scale mainly to manipulate rankings[1][2].
- AI is good at first drafts of how-to posts, FAQs and review replies, and bad at the first-hand detail that makes a local page worth ranking.
- Run every draft through a fixed editing standard: facts, first-hand detail, local specificity, compliance, byline and date.
- Never use AI to write reviews; the FTC rule bans fake reviews, including AI-generated ones[3].
What Google actually says about AI-assisted content
Google's public position has been stable since early 2023: appropriate use of AI or automation is not against its guidelines, and its systems focus on the quality of content rather than how it was produced[1]. The same guidance draws a clear line. Using automation, including AI, to generate content with the primary purpose of manipulating rankings is a violation of its spam policies[1].
The spam policies name the pattern that catches most service companies. Scaled content abuse is when many pages are generated mainly to manipulate search rankings rather than help users, and one of Google's own examples is using generative AI tools to generate many pages without adding value[2]. A roofing company that publishes a template page for each of forty nearby towns, each one a reworded copy, is the textbook case.
The practical test is not 'was AI involved?' but 'would a homeowner in this town learn something here they could not get from the next result?' If the answer is no, the page is a liability whether a person or a model wrote it.
Why generic pages no longer stand out
Small businesses adopted generative AI fast. In the U.S. Chamber of Commerce's surveys, the share of small businesses using it rose from 23% in 2023 to 40% in 2024 and 58% in 2025[4]. Your competitors can now produce a passable article on 'signs you need a new water heater' in a minute, which means a passable article is worth very little.
Source: U.S. Chamber of Commerce, 2025[4]
Show the data
| Item | Value |
|---|---|
| 2023 | 23% |
| 2024 | 40% |
| 2025 | 58% |
What still separates a page is the material only your company has: photos from real jobs, the specific equipment brands you service, the permit your county requires for a panel upgrade, the questions your office hears every week in July. AI can organize that material well. It cannot invent it, and when it tries, it produces confident errors about codes, warranties and prices that you then have to defend on the phone.
Content types: where AI helps and where it hurts
| Content type | What AI drafts well | What a person must add | Risk if left unedited |
|---|---|---|---|
| Service pages | Structure, headings, plain-language explanations | Your process, warranty terms, license number, real job photos | Generic claims that match every competitor |
| Service-area pages | Nothing useful on its own | Local housing stock, permit notes, jobs completed nearby | Doorway pages that get ignored or demoted |
| How-to and seasonal posts | Outline, first draft, checklist format | Safety limits, when to call a pro, your technicians' tips | Unsafe DIY advice, wrong code references |
| FAQ pages | Turning call notes into clear Q&A | Accurate answers on scheduling, service area, payment | Answers that contradict what the office says |
| Google Business Profile posts | Short seasonal reminders | Real offers you honor, real photos | Posts that read as filler |
| Review replies | A polite starting draft | The customer's job details, a real fix for complaints | Copy-paste replies that look automated |
The editing standard: six checks before anything goes live
Write this standard down and apply it to every AI draft, including the ones that look finished. It takes about twenty minutes per page once the habit forms.
Check every factual claim
Codes, rebates, tax credits, refrigerant rules and warranty terms change. Verify each against a primary source or your own documents, and delete anything you cannot verify.
Add first-hand detail
Insert at least three things only your company knows: a recent job, the equipment you stock, a mistake you see homeowners make, the questions customers ask most.
Run the swap test
Replace your company and town names with a competitor's. If the page still reads as true, it is not specific enough. Rewrite the parts that survived the swap.
Cut the filler
Remove openings that restate the title, paragraphs that could appear on any site, and closing summaries that repeat the page.
Do a compliance pass
Check license numbers, financing language, guarantee wording and review claims. Make sure nothing promises a result you cannot deliver.
Sign and date it
Put a real reviewer's name and role on the page and a last-reviewed date, then calendar the next review.
Briefing the tool so the first draft is closer
Most weak drafts come from weak briefs. A one-line request gets a one-size-fits-all answer. Give the tool the raw material first and ask it to organize, not invent.
- Paste your own notes: the call transcript summary, the technician's job notes, the photos' captions and the questions the customer asked.
- State the reader and the decision: 'a homeowner with a 15-year-old furnace deciding between repair and replacement this week'.
- List what must not appear: prices you have not approved, brand claims, code citations you have not checked, guarantees.
- Ask for a plain outline first, approve it, then ask for the draft section by section.
- End with an instruction to flag every sentence that states a fact it did not get from your notes, so the editor knows where to look.
Keep the brief template in a shared document so whoever writes next month starts from the same rules.
Service-area pages without doorway problems
Google defines doorway abuse as pages created to rank for specific, similar queries that lead users to intermediate pages less useful than the final destination, and it lists pages targeted at specific regions or cities that funnel users to one page as an example[2]. That does not ban area pages. It bans thin ones.
- Build an area page only where you have done real work and can show it: project photos, a short job write-up, the typical housing stock.
- Include something genuinely local, such as the permit office's process, common soil or roof types, or water hardness if it affects the equipment you install.
- Keep the number of area pages to the places you actually serve on a normal week, not every town within an hour's drive.
- Make each page useful on its own, with the phone number, hours and what happens after someone books.
Your Google Business Profile follows its own rules. A service-area business should have one profile for its central office with a designated service area, and should hide its address if it serves customers at their locations[5]. The business name should reflect the real-world name, and unnecessary information in the name is not permitted[5]. Stuffing 'Best AC Repair Springfield' into the name is a fast route to a suspension; a well-written profile description is the place for services and areas.
Example: rewriting a generic AI draft
Example (fictional): Ridgeline Roofing, a five-person crew, asked an AI tool for a page about storm damage inspections in Maple Hollow. The draft opened with a paragraph about the importance of a healthy roof, listed ten generic warning signs and closed with a call to action. It passed a grammar check and said nothing a homeowner could act on.
The office manager applied the six checks. She replaced the generic signs with the three issues the crew actually finds on Maple Hollow's 1990s subdivisions, added two photos from inspections done that spring, explained what the crew documents for an insurance adjuster, and noted the town's permit requirement for full replacements. She deleted the warning-sign list entirely. The finished page was shorter, had a reviewer's name and date, and was the only page in the area that explained what a homeowner receives after an inspection.
The point of the example is the order of work. AI produced the skeleton in minutes; the value came from twenty minutes of company knowledge added on top.
Reviews, replies and disclosure
Reviews carry more weight than most blog posts. In Pew Research Center's survey, 82% of U.S. adults said they at least sometimes read online ratings or reviews before buying something for the first time, and 40% always or almost always do[6]. That makes review replies a reasonable place to use AI drafts, as long as each reply names the job and, for complaints, the fix.
The line you cannot cross is generating reviews themselves. The FTC's final rule, announced in August 2024, bans the sale or purchase of fake reviews, including AI-generated fake reviews[3]. Do not ask staff or tools to write reviews for customers to post, and do not seed testimonials on your own site that no customer wrote.
Be thoughtful about disclosure too. Pew found 76% of U.S. adults say it is extremely or very important to be able to tell whether pictures, videos and text were made by AI or by people[7]. You do not need a label on an edited blog post, but do not present an AI-generated image as a photo of your crew's work.
A monthly content routine for a small team
- Week 1: pull the ten most common questions from call notes and texts, and pick two to answer as posts or FAQ entries.
- Week 2: have AI draft both, then run the six checks and add job photos.
- Week 3: update one existing service page with a recent project and a fresh review date.
- Week 4: draft review replies for the month, personalize each one, and check the profile's services and hours.
- Every quarter: delete or merge pages that get no visits and cannot be made specific.
For more on replies, see drafting review replies with AI. If competitors are winning with fake listings rather than content, reporting spam listings on Google Maps is the better use of your time.
Frequently asked questions
Will Google penalize my site for using AI-written content?
Not for the use of AI itself. Google says appropriate use of AI or automation is not against its guidelines, but generating content mainly to manipulate rankings, or many pages without added value, violates its spam policies.
Should I create a page for every town I serve?
Only for towns where you work regularly and can add genuinely local detail such as completed jobs, photos and permit notes. Reworded copies for every nearby town look like doorway pages.
Can I use AI to write customer reviews or testimonials?
No. The FTC's rule bans fake reviews, including AI-generated ones. Use AI only to help draft your replies to real reviews, then personalize them.
How long should editing an AI draft take?
Plan on roughly twenty minutes per page for the six checks once you have a routine. Pages about codes, rebates or safety deserve more time because errors there cost the most.
References
Google Search Central (2023). Google Search's guidance about AI-generated content (opens in a new tab) ↑
Google Search Central (2026). Spam policies for Google web search (opens in a new tab) ↑
Federal Trade Commission (2024). Federal Trade Commission Announces Final Rule Banning Fake Reviews and Testimonials (opens in a new tab) ↑
U.S. Chamber of Commerce (2025). Empowering Small Business: The Impact of Technology on U.S. Small Business (2025) (opens in a new tab) ↑
Google Business Profile Help (2026). Guidelines for representing your business on Google (opens in a new tab) ↑
Pew Research Center (2016). Online reviews and ratings (opens in a new tab) ↑
Pew Research Center (2025). How Americans View AI and Its Impact on People and Society (opens in a new tab) ↑