Skip to main content
web60

Industry News

Google AI Content Guidance: Fact-Check the Parts of Your Website You Never See

Eamon Rheinisch··12 min read
Flat illustration of a teal page shape with its corner peeling back to reveal a navy layer beneath, crossed by teal and navy diagonal strokes on an off-white background

If AI helped you write your website, you have read the words on the page. You read the headline, the About section, probably the prices twice. What you most likely have not read are the words around the page: the title Google shows in its results, the two-line summary underneath it, the descriptions attached to every photo, and the block of code that tells search engines your opening hours and your rating.

On 1 October, Google updated its guidance on AI-generated content to say you should be checking those too. "It is critical to manually factcheck and review all AI-generated content for accuracy and trustworthiness before publishing," the page reads, and that review "also applies to metadata like <title> elements, meta description elements, structured data, and alternate texts for images, which can appear in Search results" [1].

Read the end of that sentence again. Text you never look at can be the first thing your customer sees.

What Google Changed, and What It Did Not

Nothing has been banned. There is no new penalty for using AI, and if that has been keeping you up at night, our earlier piece on whether Google penalises AI-written website content covers it properly. Google judges what a page does for the reader, not which tool typed it.

The update sharpens the detail. The guidance now points to two specific sections of the Search Quality Rater Guidelines, the handbook Google's human reviewers work from: section 4.6.5 on scaled content abuse, and section 4.6.6 on main content created "with little to no effort, little to no originality, and little to no added value" [1]. Google is careful to say rater scores do not directly change rankings. Still, when someone tells you which page of the marking scheme to read, you read it.

Two practical additions sit alongside that. Online shops that send products to Google Merchant Center now face specific labelling rules for AI-generated images and product text. And Google suggests, without requiring it, that you consider telling readers how your content was made, where that makes sense for your audience [1].

The Four Places AI Writes That You Never Read

You know what most owners do after an AI tool drafts their site? They read the homepage, fix a few sentences, and deploy. Fair enough. Nobody told them about the layer underneath, and that layer is precisely what the update names.

The page title

This is the blue headline in Google's results and the words on the browser tab. AI tools write these generously. A title reading "Award-Winning Plumbing Services Near You" is a problem if you have never won an award, because it sits in front of every person who searches for you, before they have seen a single page.

The meta description

The meta description is the short summary that can appear under that headline. Google does not always use yours, but when it does, it is your sales pitch in two lines, and odds are you did not write it.

Consider a hypothetical, though the pattern behind it is common enough. A bridal boutique in Laois rebuilds its website with an AI tool over a quiet January weekend. The pages read beautifully. The meta description on the alterations page, which nobody opens, promises same-week alterations on every gown. The boutique has never offered that, and its alteration diary is booked well ahead. In spring a bride, a few weeks out from her wedding, finds the shop on Google, reads that line, and buys her dress on the strength of it. She learns the truth at the first fitting.

That is not a typo. It is a refund argument, a furious review and a story that travels, all from a sentence the owner never saw.

Flat illustration of four overlapping rounded panels shading from grey to teal, with a navy circular lens resting against the last one, on an off-white background
The text behind a page sits in layers most owners never open.

You will hear people say meta descriptions are barely worth the effort, because Google rewrites so many of them anyway. I understand the logic, and I think it misses the point. When Google does use yours, it is the only copy a searcher reads before deciding whether you are worth a click.

Image alt text

Alt text is the written description attached to each photo. Screen readers read it aloud to blind and partially sighted visitors, and Google uses it, along with the image and the surrounding page, to work out what a picture shows [4]. Historically the problem was absence: WebAIM's February 2026 analysis of the top million home pages found missing alt text on just over half of them [6], and those are big, well-resourced sites, so I would not bet on smaller ones doing better.

AI solves the empty field and creates a new problem. It guesses. A linen-blend dress becomes a "pure silk evening gown". A photo of your old premises becomes "our shop on Main Street" a year after you moved. Google's own example of good alt text is "Dalmatian puppy playing fetch", short and accurate. Its warning is just as plain: filling alt text with keywords "may cause your site to be seen as spam" [4].

Structured data

Structured data is code in the background of a page that labels facts for search engines: this is a business, these are its hours, this is a product, here is its price, here is its rating. You will never see it unless you go looking. Google's rules are blunt: "Don't mark up content that is not visible to readers of the page," and do not use structured data "to deceive or mislead users", which Google says includes marking up fake reviews [3].

So ask yourself a slightly uncomfortable question. If a tool or plugin generated the markup on your site, does it include a star rating? Where did that number come from? A rating that appears nowhere on the page, sitting in code you have never read, is exactly what those rules describe. The consequence is a manual action, which strips the page of rich results such as review stars in Google, though Google says it does not affect how the page ranks in ordinary search [3].

If You Sell Online, There Are Now Labelling Rules

This section only matters if you send products to Google through Merchant Center, for example to run Shopping ads. If that is you, the rules are specific. Images created with generative AI must carry embedded metadata saying so, the IPTC DigitalSourceType value TrainedAlgorithmicMedia [1]. Product titles and descriptions written with generative AI must be submitted in separate structured_title and structured_description attributes, flagged as AI-generated [5].

In business terms, if a tool wrote your product copy, Google expects your product feed to say so. Google's help page also warns against removing those metadata tags, which is worth knowing because some image-compression tools strip metadata to save space. Verify what yours does before you upload a catalogue.

If you never use Merchant Center, none of this touches you today.

What This Does Not Mean for a Small Site

Let me be straight about proportion. Google is not coming after a ten-page website because ChatGPT drafted the About page. Scaled content abuse, in Google's words, is "when many pages are generated for the primary purpose of manipulating search rankings" [2]: think hundreds of near-identical town pages churned out to catch searches. Sites that do that "may rank lower in results or not appear in results at all" [2]. A café with a menu, a contact page and decent photos is not in that conversation.

There is a genuine limitation to checking by hand, too. On a five-page brochure site, verifying every hidden field takes an afternoon. On a shop with a few thousand products, it does not scale, and you will be sampling rather than reviewing everything. If you run a catalogue that size with staff managing product data full time, dedicated product information software genuinely suits that workload better than editing fields one at a time in WordPress. Most local firms are nowhere near that size.

Where to Look This Week

None of this needs a developer. Set aside an afternoon and work through it in this order:

  1. Search for your own business. Type your business name into Google and read the results like a stranger would. Every title and summary there is a claim you are making.
  2. Open the SEO fields on your key pages. If your WordPress site uses an SEO plugin, each page has a box for its title and meta description. Verify each one against what you actually offer, starting with your homepage and your services pages.
  3. Read your media library. WordPress shows the alt text for every image. Correct anything inaccurate and give any meaningful photo a short, plain description.
  4. Run the Rich Results Test on your main pages. Google's free tool lists the structured data a page carries. If it reports a rating, price or opening hours you do not recognise, find out which theme or plugin put it there.
Flat illustration of a grid of rounded grey-olive tiles with a few teal tiles, one empty tile outlined in navy, and broad teal diagonal bands crossing the grid
A page-by-page check of the hidden fields takes an afternoon on a typical small business site.

Why Owning Every Field Matters

All four of those checks rest on one assumption: that you can open your own website and change things yourself. That sounds obvious until you remember how many businesses still pay an agency somewhere between €75 and €150 an hour for edits. Fixing thirty alt texts on those terms is a bill nobody budgets for, so it never happens, and the wrong descriptions sit there for years.

That is the quiet case for building your own site. What a website needs to meet Google's updated guidance is not complicated. You need to see every field, change it yourself the same afternoon you spot a problem, and never pay anyone for the privilege. WordPress gives you that access, and AI has removed the part that used to need a designer. With Web60 you describe your business and AI builds a complete WordPress site in under a minute, with hosting, SSL, nightly backups and analytics included for €60 a year, and full access to every title, description and image field from day one. The AI writes the first draft. You stay the editor.

For a wider view of where AI and WordPress are heading for business websites, our guide to AI-powered WordPress for business owners is a good next read.

The Upshot

The 1 October update did not make AI riskier to use. It put in writing something that was always true: Google treats the text behind your page as part of what you tell customers, because it is often the first thing they read. Only you know whether that text is accurate.

So pick an afternoon this week, search for your own business, and read what your website says when you are not looking. Anything you find there, you can correct yourself the same day.

Frequently Asked Questions

Does Google penalise websites that use AI-written content?

Not for using AI as such. Google's spam policy targets scaled content abuse, which it defines as many pages generated for the primary purpose of manipulating search rankings, however they were made. A small business site with a handful of accurate, useful pages drafted with AI help is not in that category. What Google does ask, in guidance updated on 1 October 2026, is that you manually fact-check everything AI produces before publishing, including the hidden text.

What is alt text and does every image on my website need it?

Alt text is a short written description attached to an image. Screen readers read it aloud to blind and partially sighted visitors, and Google uses it alongside the image itself and the surrounding page to understand what a picture shows. Any image that carries meaning, such as a product, your premises or your team, should have a short, accurate description. Purely decorative images can be left with empty alt text. Avoid stuffing it with keywords, which Google warns can make a site look like spam.

Do I have to label AI-written content on my website?

For ordinary web pages, Google suggests you consider telling readers how content was created where that makes sense, but it does not require a label. Google Merchant Center is different: if you submit products, AI-generated images must keep their AI metadata and AI-written product titles and descriptions must be submitted in separate attributes flagged as AI-generated. Other rules can apply in specific situations, so if you are unsure about your obligations, check with your solicitor.

How do I check the structured data on my website?

Paste the address of a page into Google's Rich Results Test, which lists the structured data it finds and any errors. Read what it reports in plain terms: business name, opening hours, prices, ratings. Anything you do not recognise or that is not visible on the page itself needs to be traced to the theme or plugin that added it and corrected, because Google's rules say markup must match what readers can see.

Sources

Eamon Rheinisch
Eamon RheinischSales Director, Web60

Eamon leads sales at Web60 and SmartHost, working directly with Irish business owners making the switch from cheap shared hosting to managed WordPress. With a background in enterprise technology sales — including Oracle and multiple Irish SaaS businesses — he understands the questions Irish SMEs ask before committing to a hosting platform. He writes about hosting comparisons, total cost of ownership, web design for Irish businesses, and how to evaluate what you’re actually buying.

More by Eamon Rheinisch →

Ready to get your business online?

Describe your business. AI builds your website in 60 seconds.

Build My Website Free →
Buy NowTry Free
Google AI Content Guidance: Check Your Hidden Text | Web60