On 1 October 2026, Google updated its guide to using generative AI content. It now says it is "critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing", and that this review covers titles, meta descriptions, structured data and image alt text as well as the main copy.
Many UK small business owners use ChatGPT, Claude or Gemini to keep a blog going. For them, this is the clearest statement Google has made so far. It is not a new penalty. It does make Google's expectations explicit, and it gives you a sensible standard to work to. This article explains what changed, what it means for a typical UK service business, and how to build a review habit that fits into an owner's week.
What exactly did Google change on 1 October 2026?
According to Search Engine Journal's report on 1 October 2026, the inserted text reads: "Keep in mind that generative models don't retrieve facts, but predict a likely sequence of words based on their training data. Because of this, generative AI outputs may contain inaccuracies (also known as hallucinations)." The same report notes that the phrase "This includes metadata" became "This review also applies to metadata."
Google's changelog says the page was brought in line "with information from the Search Quality Raters guidelines." PPC Land reported on 2 October 2026 that the change landed during Search Central Live Barcelona.
Google also revised its helpful content page. It now defines main content as "any part of the web page that directly helps the page achieve its purpose" and lists four attributes raters look at: effort, originality, talent or skill, and accuracy.
Is this a penalty, or just advice?
PPC Land points out that the update provides "no threshold for sampling large page sets, names no role responsible for the review and requires no record that one took place." Nobody from Google will audit your review process.
That does not make the update irrelevant. Google's guide still says that generating many pages without adding value "may violate Google's spam policy on scaled content abuse." It also says quality rater ratings "don't directly influence ranking" but inform how Google evaluates its ranking systems. Put simply, raters help Google tune the systems that decide which pages rank, and those raters are told to look for effort and accuracy. Unchecked AI copy is exactly the kind of content those systems are being tuned to rank lower.
Why does this matter more for a UK small business than a big publisher?
A national publisher with 50,000 articles can absorb a few weak ones. A Leeds electrician with 25 pages cannot. AI tools often slip in exactly the places UK owners care about:
- UK-specific rules. Models trained mostly on US text can mix up VAT with sales tax, Part P with US electrical codes, and Ofsted with US childcare licensing.
- Prices and figures. AI will happily write "a typical boiler service costs £60" with no source behind it.
- Local detail. Invented neighbourhoods, wrong postcodes, or claims to serve towns you do not cover.
- Hidden fields. Auto-generated meta descriptions and alt text that nobody reads before publishing, which Google has now named explicitly.
Which businesses need to act first?
Use this checklist to place your business:
- If you publish more than four AI-assisted posts a month, set up a written review checklist now. Volume without review is the pattern Google's scaled content abuse policy describes.
- If you work in finance, legal, health, care or childcare, treat every claim as needing a primary source. These are the topics where a wrong answer can cost readers money or harm them, and where Google gives trust the most weight.
- If you sell products online and use AI for product titles or descriptions, check your product feed. Google's guide says AI-generated product data "must be specified separately and labeled as AI-generated."
- If you use AI-generated product images in Merchant Center, confirm they carry IPTC DigitalSourceType TrainedAlgorithmicMedia metadata, which Google's guide requires.
- If you publish one or two posts a month and already edit them heavily, your main change is adding metadata to your review.
What does a proper fact-check look like for a small business?
A workable five-step routine:
- Highlight every claim. Numbers, prices, dates, regulations, place names, guarantees and named organisations.
- Verify against a primary source. gov.uk for tax and regulation, your own price list for costs, your job records for case details.
- Cut or soften what you cannot verify. "Most jobs take a day" beats a precise figure you made up.
- Add something only you know. A real job, a real customer question, a photo from site. This feeds the effort and originality attributes Google lists.
- Check the hidden fields. Title tag, meta description, structured data such as opening hours or prices, and image alt text.
If you write with Claude, building this checklist into a reusable project instruction works well. See the Claude for business set-up for how owners structure this, and the guide to planning an AI content calendar for scheduling review time into the month.
How much time and money does proper review cost?
The figures below are illustrative assumptions for a typical business, not Google data.
Take a Bristol roofing firm publishing eight AI-drafted posts a month. The owner values their time at £40 an hour.
- Review time: 30 minutes per post, including metadata, is 4 hours a month, or £160 of owner time.
- AI subscription: roughly £18 to £20 a month for a paid plan, depending on the tool.
- Total monthly content cost: roughly £180.
Now the downside. Suppose one unchecked post claims the firm offers a 15-year guarantee when the real figure is 10 years. A customer quotes that page during a dispute. Honouring the claim, or arguing it, could easily cost more than a year of review time. Losing a single £3,000 job to a visitor who spots an obvious error would cost more than 18 months of review.
The trade-off often favours fewer, better posts. Four reviewed posts with real job detail will usually do more than eight unchecked ones. If you are weighing outsourcing instead, the comparison of AI content writing services for UK small businesses covers what to ask a provider about their review process.
Does this affect visibility in AI Overviews and ChatGPT too?
Google's guidance speaks only about Search. The logic carries over, though. AI answer tools pull from pages that state clear, specific facts. A page that confidently gives the wrong VAT threshold is a liability wherever it gets quoted. Checking claims, naming sources and adding first-hand detail improves your standing with Google and makes your pages safer for AI tools to quote. For how AI search engines choose which businesses to mention, see AI visibility for small businesses.
Should you label your content as AI-assisted?
Google's helpful content page sets out a Who, How and Why framework. Under "How", it says that sharing details about the process can help readers understand the role automation played. A one-line note at the end of a post, such as "Drafted with AI, checked and edited by Sarah, owner", tells readers a real person stands behind the facts. A clear author byline, which the "Who" section also recommends, does the same job.
What should you do this week?
A concrete plan for the next five working days:
- Monday: List every page or post drafted with AI in the past six months. Sort by traffic in Google Search Console.
- Tuesday: Fact-check the top ten. Fix any wrong prices, regulations, service areas or guarantees.
- Wednesday: Review the title tag, meta description and image alt text on those same ten pages.
- Thursday: Write your review checklist using the five steps above and save it where you draft.
- Friday: Set a monthly publishing number you can review properly, even if that means cutting volume by half.
If you sell online, add a sixth task: check your product feed for AI-written titles and descriptions and confirm they are labelled as Google's guide requires.
Is this the start of a wider crackdown?
Search Engine Journal notes that Google staff, including Gary Illyes, have advised for some time that AI output should be checked by a human before it is published. The 1 October update puts that advice into Google's official guide. Owners who already review their drafts carefully have nothing new to do apart from the metadata check. Owners who publish on autopilot now have a clear, written standard they are falling short of, and it is worth fixing before Google's systems catch up with them.
Google's AI Content Fact — FAQ
Has Google banned AI-written content on small business websites?
No. Google's guidance on using generative AI content was updated on 1 October 2026. It still treats AI as a legitimate tool for research and for adding structure to original content. What it warns against is using AI to generate many pages without adding value for users, which it says may break its spam policy on scaled content abuse. The October update adds a clear instruction: it is critical to manually fact-check and review all AI-generated content for accuracy and trustworthiness before publishing. So a UK plumber or accountant can keep using ChatGPT, Claude or Gemini for drafts. The change is that a human should check every claim, price and detail before the page goes live, and Google has now written that down in its own documentation.
Does the fact-check rule apply to page titles and meta descriptions?
Yes. Search Engine Journal reported on 1 October 2026 that Google changed the wording from 'This includes metadata' to 'This review also applies to metadata', and the guide lists title elements, meta descriptions, structured data and image alt text. This matters for small businesses because many website builders and SEO plugins now fill these fields in with AI. A meta description that promises same-day callouts across Manchester when you only cover two postcodes is an accuracy problem, even though few visitors see the field on your page itself. It also appears in search results, so it shapes the first impression. Treat every AI-filled field as a draft, not a finished item.
Will Google penalise my site if I do not fact-check AI content?
The update itself introduces no new penalty or enforcement mechanism. PPC Land noted on 2 October 2026 that the guidance sets no threshold for sampling large page sets, names no role responsible for the review and requires no record that one took place. The risk is indirect but real. Inaccurate pages perform poorly with visitors, and Google says its quality raters judge effort, originality, talent or skill, and accuracy. Those ratings do not directly change rankings, but Google says they inform how it evaluates its ranking systems. Publishing large volumes of unchecked AI pages also moves you towards what its spam policy calls scaled content abuse.
Do I need to tell readers that a blog post was written with AI?
Google does not require a label on ordinary blog posts. Its helpful content page describes a Who, How and Why approach, where sharing how content was created, including any use of automation, can help readers understand it. A short note such as 'drafted with AI, checked and edited by the owner' is a reasonable choice for a UK service business. The firmer rules apply to ecommerce. Google's guide says that for Merchant Center, AI-generated product images must carry IPTC DigitalSourceType TrainedAlgorithmicMedia metadata, and AI-generated product titles and descriptions must be specified separately and labelled as AI-generated in your product data.
How long should fact-checking an AI blog post take?
As a rough guide, a careful owner review of a 1,000 word service article often takes 20 to 40 minutes, depending on how many prices, legal points and local details it contains. That is our own estimate, not a Google figure. The fastest method is to mark every checkable claim in the draft, such as numbers, dates, regulations, place names and service promises, and verify each one against a primary source or your own records. Then check the title tag, meta description and image alt text. Posts on tax, finance, health or legal topics deserve more time, because mistakes there can cost readers money or harm them, and Google's guidance gives extra weight to trust on those topics.
