AI content and SEO: which jobs to hand AI, and what to disclose
Short answer Give AI the drafting: outlines, first drafts, titles, meta descriptions, alt text and rewrites of old pages. Keep the facts, prices, photos, experience and final approval with a person. Google ranks the page, not the tool, but asks you to fact-check everything. Never let AI write reviews, and keep customer data out of consumer chatbots.
Want help applying this to your business? Talk it through on a free 30-minute call.
Book a callWhich SEO content jobs should I hand to AI?
The ones where a mistake is easy to spot and cheap to fix: outlines, first drafts, short page labels and rewrites. Keep a person on anything a customer will act on, such as a price, a rule, a claim about your work, or a review.
Google's own line on where AI helps:
"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)
| SEO job | Hand to AI? | What the person still does |
|---|---|---|
| Brainstorm topics from customer questions | Yes | Pick the ones you actually sell and can answer |
| Keyword ideas | Ideas only | Check real search numbers in a keyword tool; chatbots guess them |
| Page outline | Yes | Reorder it around what customers ask first |
| First draft of a service page or post | Yes | Add prices, process, photos, local detail; cut filler |
| Title tag, meta description, image alt text | Yes | Read every one: Google says the same review applies |
| Schema markup (code that labels your page for search engines) | Draft only | Make every value true, then test it |
| Refresh an old page | Yes | Confirm what changed in your business |
| Translate a page | Draft only | A fluent person checks it |
| Facts, prices, laws, medical or legal claims | No | A person writes or verifies each one |
| Customer reviews and testimonials | Never | Ask real customers |
The metadata row comes straight from Google: its guidance says title elements, "meta description elements, structured data, and alternate texts for images" need the same review as the page (Google Search Central). And its spam policy counts copying other sites' content through automated "translating" as mass-produced spam (Google spam policies), so translate your own pages, checked by a fluent person.
Will AI-written content hurt my rankings?
Not by itself. Google judges the page, not who or what typed it:
"Using AI doesn't give content any special gains. It's just content. If it is useful, helpful, original, and satisfies aspects of E-E-A-T, it might do well in Search." (Google Search Central Blog, February 2023)
E-E-A-T stands for experience, expertise, authoritativeness and trustworthiness: signs that a real, qualified business stands behind the page. What does get punished is volume without value. Google's spam policy calls it scaled content abuse and gives "using generative AI tools or other similar tools to generate many pages without adding value for users" as an example (Google spam policies). The full policy is in Does Google penalize AI content?
So the risk isn't AI. It's publishing 50 pages nobody needed because AI made them free.
Why does a person have to check every AI draft?
Because AI tools don't look facts up; they write what sounds likely. Google puts it plainly:
"Keep in mind that generative models don't retrieve facts, but predict a likely sequence of words based on their training data." (Google Search Central)
And:
"It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing." (Google Search Central)
For a service business the costly errors are specific: a wrong permit rule, a price you don't charge, a service you don't offer, a city you don't cover. A five-minute checklist catches most of them:
- Every number, price and date matches what you'd say on the phone.
- Every rule or law is checked against the official source.
- The page names only services and areas you really cover.
- At least one thing on it only your business could write: a photo, a step from your own jobs, a common mistake you see.
- You'd sign your name under it.
Point 4 is what Google's helpful-content checklist looks for: content that shows "first-hand expertise and a depth of knowledge" (Google Search Central). AI can't supply it. More in E-E-A-T for a small business.
Do I have to tell readers I used AI?
For ordinary web pages, no rule in Google Search requires a label. Google's advice is to disclose where a reader would wonder:
"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 Blog)
In practice:
- A service page or how-to post a person checked: a real name and role on it does more than an AI label. Google suggests "accurate author bylines when readers would reasonably expect it" (Google Search Central Blog).
- Content mostly made by automation, such as a large set of generated pages: say how it was made. Google says to consider "providing more background information on how automation was used" (Google Search Central).
- Don't list AI as the author. Google says an AI byline "is probably not the best way" to tell readers AI was involved (Google Search Central Blog).
- Selling products through Google Merchant Center: labeling is required, not optional. "AI-generated images must contain metadata using the IPTC DigitalSourceType TrainedAlgorithmicMedia metadata," and AI-written product titles and descriptions must be "labeled as AI-generated" (Google Search Central).
One honesty line that isn't in any Google rule but matters to customers: don't present AI images as photos of your own jobs, team or premises. If a picture is generated, don't caption it as a real project.
What are the ethical lines for AI content?
Four lines cover most of the risk for a small business.
1. No AI-written reviews. In the US, the FTC's rule on fake reviews, announced August 14, 2024, bans writing, buying or selling reviews from people who don't exist or never used the business, and its announcement names "AI-generated fake reviews" as an example. It lets the FTC seek civil penalties (fines) against knowing violators (FTC). In Canada, the Competition Bureau has warned about chatbots used to "spread fake reviews that sound and look like real people wrote them" (Competition Bureau, March 2024). See rules for asking for reviews.
2. Keep customer data out of consumer chatbots. Pasting a client's name, address or job details into a free AI tool shares it with that company. OpenAI says that for individual services like ChatGPT "we may use your content to train our models" unless you opt out, while for business plans and its API, "by default, we do not train on any inputs or outputs" (OpenAI). Check your tool's setting, or remove names before you paste.
3. Don't mass-produce to game search. Google's helpful-content checklist asks: "Are you using extensive automation to produce content on many topics?" (Google Search Central). A yes is a warning sign, and it's also a poor deal for readers.
4. Know what you own. In the US, AI output alone isn't protected by copyright. The Copyright Office concluded that "copyright does not extend to purely AI-generated material," and that "prompts alone do not provide sufficient human control" to make you the author. Your own writing and edits stay protected even when a page includes AI text (U.S. Copyright Office, January 2025). The more of yourself you add, the more of the page is yours.
Example
A roofer with ten service pages
He asks AI for an outline and draft of each page, then spends 20 minutes per page adding his warranty terms, his price range ($8,000 to $15,000 for a typical re-roof) and two job photos. He removes a paragraph claiming a "lifetime guarantee" he doesn't offer. Ten pages, about four hours of his time, all accurate.
A cleaning company tempted by city pages
AI can write 60 "house cleaning in [town]" pages in an hour. The owner serves 12 towns, so she publishes 12, each with the team that covers it and local parking or building notes. The other 48 would have been near-copies, the pattern Google's spam policy describes.
A bookkeeper with client questions
She wants AI to turn real client emails into an FAQ page. She deletes names and figures first, uses her firm's business AI plan, and checks each tax rule against the official source before publishing.
FAQ
Can AI do keyword research for my business?
It can suggest ideas, but chatbots don't have real search counts and often guess them. Check numbers in a keyword tool; see keyword research for a local business.
Should I run AI drafts through a plagiarism checker?
It's a reasonable spot check, since AI can echo phrases from its training data. A page full of your own prices, photos and process rarely matches anything else.
Can I use AI to update old blog posts?
Yes, it's one of the safer uses. Confirm every changed fact; see when to update, merge or delete old pages.
How do I make AI drafts sound like my business?
Paste in two or three pages you wrote yourself and ask it to match that tone. Then read the draft aloud and fix anything you'd never say.
How much AI content is too much?
Google sets no percentage. The test is whether each page helps a customer and was checked by someone who knows the work.
Different for you?
- Worried a past AI page got your site penalized: Google manual actions
- Unsure what AI can do across all of SEO, not just writing: Can AI do SEO?
- Deciding whether a blog is worth it at all: Do blogs help SEO?
- You want a team to draft, check and publish pages: SEO agent