No. Google does not penalise a page because AI helped write it; its own guidance says appropriate use of AI or automation is not against its rules. What it demotes is low-value content produced at scale mainly to manipulate rankings, which its spam policies call scaled content abuse, however it was made. For most small business sites, the real risk is pages that never rank.
That distinction matters because the fix is different. A genuine penalty is rare for a site with a few dozen pages. The common problem is quieter: an unedited AI draft says what every competitor's page already says, so Google has no reason to show yours. An optician's "What happens in an eye test?" page produced in thirty seconds competes with thousands of near-identical ones, and loses.
What Google's published rules say about AI-written pages
Google has been unusually direct on this. In February 2023 it published a statement saying it rewards helpful, original content however it is produced, and that using AI or automation appropriately does not break its guidelines. Its documentation on using generative AI content, last updated in December 2025, goes further: it describes AI as useful for researching a topic and adding structure to original writing, and warns that generating many pages without adding value for users may breach the spam policy on scaled content abuse.
That spam policy is the one to understand. Google defines scaled content abuse as generating many pages primarily to manipulate search rankings rather than help people. Its examples include using generative AI to produce many pages without adding value, and scraping or automatically transforming other content (rewording with synonyms, translating, or otherwise disguising it) where little value is added. The policy is about intent and value, not the tool. A person typing 300 thin pages by hand breaks it just as surely.
Google's guide to creating helpful, people-first content adds three questions it wants publishers to ask of every page: who made it, how, and why. The "how" question asks whether the use of automation, including AI, is clear to visitors where they would reasonably expect to know. The "why" question is the sharp one: if the main reason a page exists is to attract search traffic rather than to help the people who land on it, Google treats that as a warning sign.
The human side of the system points the same way. Google's search quality rater guidelines, the manual it gives to the people who assess sample results, were updated in January 2025 to tell raters to give the lowest rating when nearly all of a page's main content is copied, paraphrased or AI-generated with little effort or added value. Raters don't move individual pages up or down; their scores help Google judge whether its ranking systems are working. But it shows what the systems are being tuned to find.
Penalty, demotion or simply invisible: three different problems
Owners often call any traffic drop a "penalty". It helps to separate the three situations, because each one shows up differently and needs a different response.
| Situation | How you find out | Typical scope | What to do |
|---|---|---|---|
| Manual action | A notice in Search Console under Security & Manual Actions, then Manual actions, plus an email to verified owners | Part of the site or all of it | Fix every issue named, then submit a reconsideration request explaining what changed |
| Algorithmic demotion | No notice; a sharp drop that lines up with a spam or core update date | Often site-wide | Remove or improve the low-value pages; recovery comes as systems reassess, often months later |
| Never ranked, or faded | No notice; pages get few or no impressions from the start | Page by page | Rewrite with first-hand detail, merge overlapping pages, or retire them |
For a nail salon, pet shop or optician with 20 to 80 pages, the third row is by far the most likely. It isn't a punishment. It is Google choosing a more useful page over yours, which you can change.
Where AI use on a small business site crosses the line
The line isn't "AI wrote a sentence". It's volume without value. Three illustrative cases show where owners drift over it without meaning to.
The pet shop that reworded its supplier feed
Picture an independent pet shop that imports 400 products from a wholesaler. The wholesaler's descriptions are on dozens of other retailers' sites, so the owner asks an AI tool to "rewrite each description so it's unique". The results are technically different sentences carrying exactly the same information. That is close to the "automated transformations like synonymising" example in Google's policy. The better route is to rewrite the 40 products that sell most, adding what only the shop knows: which dog breeds the harness actually fits, which cat food fussy eaters in the shop's own customer base switched to, what the bag weighs in real life. Leave the rest as honest supplier copy and let the category pages do the work. Our guide to product descriptions that sell covers that rewrite in detail.
The nail salon with a page for every nearby town
A salon owner generates 25 pages, one per neighbouring town, each saying "looking for gel nails in [town]? Our friendly team…" with the name swapped. Google's spam policies already treat these as doorway pages, and AI just makes them faster to produce. One strong page about the salon, its services and where clients travel from does more than 25 clones.
The garden centre that published 150 care guides in a weekend
Here the pages were individually readable, but nobody checked them. Several recommended planting frost-tender bedding weeks too early for the centre's climate, and two contradicted the advice staff give at the till. Scale plus no review is the pattern Google targets, and wrong advice damages trust with customers long before any algorithm notices.
Compare that with uses that are plainly fine: drafting a first version you then rewrite, restructuring a messy page into clear headings, writing meta descriptions, turning a staff member's spoken explanation into tidy text, or summarising 50 real customer questions into one FAQ page.
An optician's service pages, before and after editing
An illustration of the editing that makes the difference. A two-practitioner optician wants 12 service pages: eye tests, children's eye tests, contact lens fittings, dry eye clinics, frame styling and so on. The owner drafts each one with AI in about five minutes. Here is part of the eye test draft as the tool returned it:
Regular eye tests are an essential part of maintaining your overall health. Our experienced optometrists use state-of-the-art equipment to assess your vision and detect potential issues early. We pride ourselves on delivering a friendly, professional service tailored to your needs.
Nothing in that paragraph is false, and nothing in it is specific. Every optician's site says the same. After 20 minutes with the practice manager, the same section reads:
An eye test here takes about 30 minutes. We check how clearly you see at distance and for reading, measure the pressure in your eyes with a puff-free test, and photograph the back of your eye so we can compare it next time. Bring your current glasses and a list of any medicines. If you drive for work, tell us when you book and we'll leave time to talk about glare at night.
The second version answers what a nervous first-time patient actually wonders: how long, what happens, what to bring. It contains details a competitor's page can't copy because they describe this practice. The time cost is the useful sum here: 12 pages at 5 minutes of drafting plus about 20 minutes of editing is roughly 5 hours in total, against perhaps 12 to 15 hours writing from a blank page. AI saved most of the typing; the owner still supplied the substance.
A value check before any AI draft goes live
Run each draft through these questions. If it fails three or more, don't publish it yet.
| Question | What a pass looks like |
|---|---|
| Does it contain at least three facts only your business could state? | Prices, timings, process steps, named services, your policy wording |
| Would a customer learn something they couldn't get from the first search result? | A practical detail, example or answer to a follow-up question |
| Has someone who does the work read it? | A named person checked facts and advice |
| Is there already a page on your site covering this? | No overlap, or the old page is merged and redirected |
| Does the opening answer the question the page is named after? | First two sentences give the answer, not scene-setting |
| Are claims supportable? | No "best in town", no invented statistics, no guarantees you can't evidence |
| Would you be happy if a regular customer read it aloud to you? | It sounds like your business, not a brochure template |
| Is it written for a person rather than a keyword? | The search phrase appears naturally, not in every paragraph |
| Do images show your real premises, stock or work where it matters? | Stock or AI images only where authenticity isn't the point |
| Would you still publish it if search engines didn't exist? | Yes, because customers ask this question |
The last question is Google's "why" test in plain words. It's the quickest way to catch a page that exists only to rank.
Prompting for a draft that isn't everyone else's
Most generic AI copy comes from generic prompts. Give the tool your raw material first, and ask it to work only from that. A template you can adapt:
You are helping a small [business type] write a web page titled "[page title]".
Use ONLY the facts in the notes below. If something a customer would want to
know is missing, list it as a question for me at the end instead of guessing.
Audience: [who reads this, what they're worried about]
Our notes (from staff, in their words):
- [how long it takes, what happens, what it costs or where prices are]
- [what to bring or prepare]
- [the questions customers ask most often at the counter]
- [one thing we do differently, with a real example]
Write 350-500 words. Plain British English. Short paragraphs. No phrases like
"state-of-the-art", "tailored to your needs" or "we pride ourselves".
Start with a two-sentence answer to the page title.Run for a dry cleaner's "Wedding dress cleaning" page, with notes about a three-week turnaround, hand-finishing of beading and a boxed-storage option, an illustrative return looked like this:
We clean most wedding dresses in about three weeks, and every one is finished by hand. Beading, sequins and lace are checked before cleaning and covered or stitched where they're loose, so nothing comes back missing... Questions for you: Do you charge extra for trains over a certain length? Can customers collect sooner for a fee? What happens if a stain can't be removed?
What to fix: the draft said "most" dresses without saying which ones take longer, so the owner added a line about silk and heavily beaded gowns. The three questions at the end were the most useful part, because they pointed to gaps a customer would notice. Answering them turned the page from adequate to genuinely helpful. The tone work behind prompts like this is covered in why AI marketing copy sounds generic.
Disclosure, bylines and AI images
Google doesn't require an "AI-written" label on ordinary business pages. Its guidance suggests explaining how content was made where readers would reasonably expect to know, and recommends bylines where one would be expected. For a service business, a practical approach is a named author or reviewer on advice pages ("Reviewed by the practice's lead optometrist") and no label on routine pages such as opening hours or a services list. Nobody expects a byline on a returns policy.
Two areas have firmer rules. For AI-generated images, Google asks sites to keep the IPTC DigitalSourceType metadata value TrainedAlgorithmicMedia that many image tools embed, so don't strip it out when you compress files. For online shops using Google Merchant Center, AI-generated product data such as titles and descriptions must be supplied separately and labelled as AI-generated. If a garden centre uses AI to write its product feed, that labelling step belongs in the upload routine.
Whether to tell customers more broadly is a brand decision rather than an SEO one; telling customers you use AI works through the options.
How to tell whether an update has hit your site
If traffic falls, check the evidence before blaming AI. The steps take about 30 minutes:
- In Search Console, open Security & Manual Actions, then Manual actions. "No issues detected" rules out a manual penalty.
- Open the Performance report, set the date range to the last 16 months, and look for the exact day clicks dropped.
- Compare that day with Google's Search Status Dashboard. In 2026 Google has run spam updates starting on 24 March, 24 June, 18 August and 24 September, plus core updates starting on 27 March and 21 May.
- Filter the Performance report by page. If the drop is concentrated on the pages you produced quickly and in bulk, that points to a quality reassessment. If it's spread evenly, look elsewhere.
- Compare the same weeks last year. Seasonal businesses often mistake a normal dip for a penalty.
A realistic mistake shows why step five matters. A dry cleaner saw clicks fall by about a third in late summer and assumed Google had punished the AI-drafted pages added in spring. The year-on-year comparison showed an almost identical dip the previous August, when office workers were on holiday and fewer suits came in. The real problem, found in step four, was different: a site redesign had broken the redirect from the old "alterations" page, which had quietly dropped out of the index. Fixing one redirect recovered more than any rewrite would have.
If a drop does line up with a spam update and sits on bulk-produced pages, improve or remove those pages and give it time. Google has said sites affected by its systems can recover, but the reassessment isn't instant and there is no form to request it for algorithmic changes.
A publishing pace that stays safe for a small site
There is no published page limit, and a quota would miss the point. What protects a small site is a routine that keeps a person in charge of every page. A workable rhythm for most owners:
- Two to four new pages a month, each starting from real questions customers ask, drafted with AI and finished by someone who does the work.
- One old page improved a month. Update prices, add a photo of the real thing, fold in a new customer question.
- A quarterly prune. Pages with zero impressions after six months get merged, improved or removed.
- One owner for accuracy. Even in a three-person business, one person signs off anything that gives advice.
The quick sum for a barber shop: four new pages a month at about 45 minutes each, including editing, is three hours a month. That produces around 48 solid pages a year, which is plenty for a local business. The owner who produces 200 pages in a weekend has created 200 pages to maintain, most of which nobody will find, and has handed Google the exact pattern its spam policy describes. If you want the broader workflow, from topic choice to publishing, see writing your core website copy with AI and build a brand voice guide first so every draft starts closer to how you speak.
Questions owners ask about AI content and Google
Can Google detect AI-written text?
Google hasn't said it ranks pages on whether a detector thinks they are AI-written, and its guidance judges quality rather than origin. Detection matters less than you'd think: pages that are generic, repetitive or published in bulk look low-value to Google's systems whether a person or a model wrote them. Spend your effort on adding real detail, not on disguising the tool.
Should I delete AI pages I've already published?
Not automatically. Check each page in Search Console: if it gets impressions and answers a real question, improve it with your own detail. If it gets no impressions after several months, duplicates another page or says nothing specific to your business, either merge it into a stronger page with a redirect or remove it. Fewer, better pages usually serve a small site well.
Is AI-translated content treated differently?
Google's spam policy lists automated translation among the transformations that can count as scaled content abuse when little value is added. A translated page that a fluent person has checked, and that serves customers who read that language, is a normal page. Machine-translating your whole site into many languages without review, just to catch more searches, is the risky version.
Do AI-written meta descriptions or alt text cause problems?
No, as long as they are accurate. Meta descriptions and image alt text are short, factual fields where AI is a good time-saver. Check that each description matches its page and that alt text describes the actual image, not keywords you'd like to rank for. Keyword-stuffed alt text helps nobody, including visitors using screen readers.
Further reads
- How to Write Website Articles With AI That Still Rank on Google — The full workflow for AI-assisted articles that still earn rankings.
- How to Fact-Check AI Marketing Copy Before It Goes Live — A checking routine for claims, prices and figures in AI drafts.
- What Is Generative Engine Optimisation and Does Your Business Need It? — How the same pages get cited by AI answer engines.
- How to Audit Your Website's SEO With AI in One Afternoon — Find thin or duplicate pages on your site before Google does.
- How to Do Keyword Research With AI for a Small Business Website — Choose page topics people actually search for.
- How to Choose an SEO Agency That Understands AI Search — What to ask an SEO agency about AI content and AI search.
- How Wedding Businesses Get Found When Couples Ask AI for Suppliers — How AI assistants build supplier shortlists for couples, a test to see what they say about you, and the five fixes that get a wedding business named.
- Is It Cheaper to Use AI or a Freelancer for Marketing Content? — A per-piece cost comparison of AI and freelance writers for marketing content, counting your own editing time, with three worked cases.
- 12 AI Marketing Mistakes Small Businesses Make (and the Fixes) — Twelve ways AI marketing goes wrong in small businesses, how each one shows up, and the specific fix and check that stops it happening again.
- What AI Can and Can't Do for Your Local SEO — Which local SEO jobs AI does well, which it can't touch, and an afternoon's worked example for a dry cleaner, mapped against Google's own ranking factors.
- How Much Does Human Editing of AI Content Cost? — What editors charge to fix AI drafts, from a quick proofread to a full fact-check, with a costed monthly budget you can adapt.
- How to Audit Your Website for AI Content That Needs Fixing — A content audit for sites with AI-drafted pages: inventory, leftover searches, a five-check score, keep-fix-merge decisions and a translation agency's numbers.
- AI Website Builder or Web Designer: What Each Really Costs — Builder prices for Wix, Framer, Durable and Hostinger, what designers charge according to Clutch, and a three-year cost that counts your own hours.
- AI Tools and AI Development: The Complete 2026 Guide — the AI hub, including every tutorial in the AI-for-business series.
Sources: Google Search Central documentation on generative AI content, spam policies (scaled content abuse, updated August 2026) and creating helpful, people-first content; Google Search Central statement on AI-generated content (February 2023); Google Search Status Dashboard ranking update history; Google Merchant Center guidance on AI-generated product data.