Google's New AI Training Opt-Out: What Cyprus Websites Should Do

On 30 October 2026 Google announced that it will support a shared "AI training" switch inside robots.txt, the small text file every website already has. One line will tell Google, and the other AI companies that join the scheme, whether your site's text and photos may be used to train their models. Google aims to support it, together with other providers, during 2027.

Most Cyprus business owners have never opened their robots.txt. Here is what it is, what the new switch does and does not do, and what to decide before your developer asks.

What robots.txt is, in thirty seconds

Type your domain followed by /robots.txt into a browser and you will see a few lines of plain text. It is a notice board for automated visitors. Google's crawler reads it before visiting your pages, and so do the crawlers of other search engines and AI companies. The lines say which parts of the site they may read.

It is a request, not a lock. Well-behaved companies honour it; anyone else can still fetch your pages like any visitor.

What Google announced

Until now, keeping a site out of AI training meant naming each company's crawler separately. Google has offered its own control, called Google-Extended, since 2023. Other companies use other names, and keeping the list current was a job nobody did.

The new approach, developed at the Internet Engineering Task Force, the body that sets web standards, uses one shared "policy token". Write it once and every participating AI provider is expected to read it the same way. If your file opts out with the shared token and says nothing about Google-Extended, Google will treat that as a Google-Extended opt-out too.

The old controls keep working, so you can state a broad choice and still make an exception for one company. Google notes the plan fits the EU AI Act code of practice, which treats robots.txt as the main way a website states its preferences in a form machines can read. The exact wording is still a draft, so there is nothing to type yet.

Training is not the same as search

AI companies do three different things with your website:

• Index it for search results. Ordinary Google. Block this and you disappear from search.
• Read it live to answer a question. AI Overviews, Gemini and ChatGPT search look at pages when someone asks. This is where mentions and clicks come from.
• Train a model on it. Your text becomes part of what the model learned, with no link back to you.

The new switch is about the third, but the line is not clean. Google's post says that sites which specifically allow Google-Extended are eligible to appear, and receive clicks, in answers from the Gemini apps. AI Overviews in Google Search follow the ordinary search rules and are not affected. So opting out of training may cost you a mention in Gemini while leaving Google Search alone.

Who should opt out

If your content is the product, opting out is reasonable. Photographers, news sites, authors and anyone selling guides or courses are giving away their stock when a model trains on it, and the switch lets them say no in one line.

If your website exists to be found, think twice. A hotel in Protaras or a dental clinic in Nicosia writes its pages so that people, and increasingly AI assistants, recommend it. For that business, being quoted is the point, and a training opt-out that removes you from Gemini answers works against it. We covered how assistants pick local businesses in our ChatGPT guide and how to get cited in AI Overviews.

Either way, it is the owner's decision, not the developer's, and it belongs on the website ownership checklist.

What to do this month

Open your robots.txt and read it. We regularly find a "Disallow: /" line left over from a staging site, quietly blocking Google from everything. Make sure it does not block Googlebot or your service pages.

Write your decision in one sentence, "our content is marketing, keep it open" or "our content is the product, opt out of training", and give it to whoever manages the site.

If you want out now, you need not wait for 2027. Ask your developer to add the Google-Extended opt-out and the matching lines for the other main AI crawlers today. The shared token goes on top when it is published, and the named lines stay valid beside it. Check the file again after any redesign, because rebuilds are when robots.txt gets overwritten.

What the switch will not do

It will not pull your content out of models that already trained on it, stop a company that ignores robots.txt, or stop a person copying your photos by hand. Your copyright is unchanged either way. What it offers is a clear, cheap way to state a preference that the main AI companies have said they will respect, and that is worth a minute of your time.

Not sure which side your business falls on? Book an AI consultation or send us a message and we will check your robots.txt with you.