Google says a program reading your site is the whole application

The sentence Google put on a live help page

Google's own wording, dated 18 September: "you don't need to provide additional data beyond what's accessible through web crawling". That is the qualifying rule for a new result box in search. The box is called the supplier unit, and it shows one business directly.

Crawling is Google sending an automatic program out to open your webpage and read whatever comes back. So the whole application is a program visiting your site. Feeds of data are an optional extra on top. Both the supplier-unit page and Google's public changelog for search carry the same 18 September date.

The box only shows for people searching from Europe. Google redesigned these results to satisfy an EU law called the Digital Markets Act, which put it on a 60-day clock from a European Commission order on 23 July.

Google has announced no US rollout, and the reporters who read those pages on 18 September found no such statement either. Do not go looking for this box on your phone tonight. You will not find it.

Crawling decides what an assistant can say about a US business, too. A program opens the page, reads what comes back, and works with that. ChatGPT, Claude, Perplexity and Gemini have been answering questions about local businesses that same way for years, with no directory and no European law anywhere in it. Google has now put the rule on a live product page in plain words.

There is no form to send and no account to keep updated. Whatever sits on the page the day a program reads it is the whole application.

What a program gets back when it opens your page

Plenty of small business sites show their hours, their prices and the area they cover on screen while keeping those words out of reach of a program. The words arrive through JavaScript, which is code that builds part of a page after it first loads, the way a live chat box or a booking calendar does.

A booking calendar is the clearest case. Your hours are stored in the calendar company's system, and the page pulls them in a moment after it opens (a second or two, which is why you never notice). Your customer sees Monday 8 to 5. A program reading the code the page handed back gets an empty box with a loading spinner, and never sees the hours at all.

Somebody measured how often that happens. Vercel and MERJ published a study of website reading programs on 17 December 2024, covering the ones OpenAI, Anthropic, Meta, ByteDance and Perplexity send out. At the time of that study, none of them ran JavaScript at all.

In that December 2024 study, ChatGPT's program fetched JavaScript files in 11.50% of its requests and Claude's in 23.84%. Neither one ran them. Whatever those files were going to put on the page stayed out of reach of both companies.

Two programs in that December 2024 study did run JavaScript. Google's Gemini uses the same setup as Googlebot and renders the page, and Apple's crawler puts it through a browser.

That study is 21 months old now and it was published long before this week's change. The gap it measured may well have closed since for some of these companies, and nobody has published a fresh measurement either way.

The ten-minute check on your own pages

You can do this from the couch with a laptop, and you do not need whoever built the site.

1. Open your services page and write down the facts a customer would ask for before calling: your hours, your prices, your phone number, your address, and the area you cover.
2. In Chrome, Edge or Firefox, right-click anywhere on the page and choose View Page Source. A window of code opens in a new tab. Safari hides this until you turn on the Develop menu in Settings, under Advanced.
3. Press Ctrl+F in that window, or Cmd+F on a Mac, and search for the bare number, like 8 rather than 8am or 8:00, because the code can write a time differently from the way the page shows it. Do the same for one of your prices and for your phone number.
4. Anything you cannot find in that window is a fact no program can read off the page.
5. Send that list to whoever handles your site and ask for those facts to be typed into the page as ordinary text.

I have run this check on a lot of small business pages, and the hours are what go missing most often. The usual culprits turn up fast: hours inside a booking widget, prices inside a photo of a menu, a phone number that only appears after a chat bubble loads, and a price list sitting in a linked PDF.

Dental practices and medspas usually trip on the price list. For HVAC shops and home-care agencies it tends to be the area they cover, which a map widget often draws on screen while the page never names the towns in words.

The fix is usually smaller than it sounds. A services page carrying a short list of prices in ordinary text, and an hours block typed out line by line, covers most of what a customer wants to know before calling. Your web person can add both in under an hour, and the booking widget can stay exactly where it is.

How much weight to put on this

Google has said nothing about how often these boxes show for local searches, or about which directories it has approved so far. The trade coverage on 18 September also found no word from the European Commission on whether the redesign satisfies its order.

A US version would arrive with an announcement from Google, and none has come. Anybody selling you preparation for it this month is selling you a guess.

The redesign does put one requirement on directories. A directory that wants into the companion box, the one showing a directory's picks, has to be approved by Google and send it a feed of listings. The approval and the feed are the directory's job.

What a US owner can use is the rule itself. Google wrote down that reading the page is the whole test, and the assistants your customers already ask have been applying that same test every day for years. Your hours and prices are either typed into the page as words, or the program reading it comes away without them.