The robots meta tag is one row, not the whole control page
Google's page on what you share with Search was last updated 21 October 2026. The robots meta tag is one row on it, grouped with data-nosnippet and the X-Robots-Tag. A separate row is the Search generative AI control. This is a reading of that page. It is not a promise that a setting has already cleared a site.
Seven rows, and where the robots meta tag sits#
The robots meta tag is not the whole page. The controls page, last updated 21 October 2026, lists seven methods: remove the content, password-protect files, the noindex rule, disallow crawling with robots.txt, the robots meta tag with data-nosnippet and the X-Robots-Tag, the Search generative AI control, and an opt-out of specific Google properties. A quiet quarry is the gap between one tag and seven rows.
The robots meta specification, last updated 24 March 2026, is the detail page. It says these settings are read only if crawlers can reach the page. Data nosnippet is a text-level attribute, not a site switch. The x-robots-tag is the HTTP header form, useful when you want one rule across a file type. The noindex page says the page must not be blocked in robots.txt, or the crawler never sees the tag.
| Item | Value |
|---|---|
| 2026-03-24 | Meta spec |
| 2026-10-15 | Crawler list |
| 2026-10-21 | Controls page |
The AI row, and what it does not do#
Ai overviews are named on the Search generative AI control help page, with AI Mode and generative AI features in Discover. Exclude means links and content from the site are not used as an input for those features. The same page says the control does not affect AI training. Training is the Google-Extended token, on a crawler page updated 15 October 2026. A moonlit muddle is treating those two switches as one.
The AI Overviews help page describes the feature for searchers. It is not the site control. The removal page says a lasting removal is to change the page, password-protect it, or add noindex. It says not to use robots.txt as the way to hide a page. The noindex tag, on that reading, is a search result rule, not a lock on the file.
The Search Central blog and the documentation updates page and the what's new page did not, in the text read here, announce a new control on 21 October. The change is on the controls page itself. A hedgerow of blog headlines is not a substitute for the updated table.
A second pass for the room. The robots meta tag is one row. The noindex tag is a different row, and it fails if robots.txt hides the page. Data nosnippet trims a passage. It does not remove a URL.
The x-robots-tag is the header, not a second philosophy. Ai overviews are a feature the AI control can exclude a site from, on the help page's list. A careful count is seven methods. A plain page can hold them. A wary watch stops anyone calling the AI row a training ban.
A quiet check is to name the page before the verb. The quiet den of a block tag is not a ranking. The trail back to the controls page is the whole job. A prowl through seven rows beats a single triumphant sentence. A waffle of one tag presented as the whole page is the defect. If a sentence cannot point at one of these URLs, it does not ship.
What to say when someone asks for a ranking from a block#
Say no, and say why. The robots meta tag does not carry a lift in any page read here. The noindex tag removes a URL from results only after the crawler sees it. Data nosnippet trims a passage. The x-robots-tag is the header form of the same family. Ai overviews can be excluded as an input, on the help page, without touching training. A careful count of those jobs is the brief. A plain page can hold it.
Then date the documents. 21 October 2026 for the overview. 24 March 2026 for the tag grammar. 15 October 2026 for the crawler list that holds Google-Extended. A wary watch catches a sentence that treats those three dates as one launch. A quiet check is to ask which URL the verb came from. The quiet den of an undated claim is where this note goes wrong.
If the room wants a chart, the honest chart is a count of seven methods, captioned as a row count, not a score. The other picture is a timeline of the three documentation dates. There is no percentage of rankings saved. Do not draw one.
A lantern over a blank score box is the picture. The thicket is a promised place in the results. The trail ends at the controls page. A prowl past that page, into a lift study, is not this update.
How to brief the robots meta tag without a ranking promise#
Write seven methods, dated 21 October 2026. Write the specification date, 24 March, when you need the tag grammar. Write Google-Extended, 15 October, when someone asks about training. Do not tell a client the block will lift a ranking. The pages read here do not say that. Say so twice now.
If the job is a search plan that has to sit beside this page, the measurement habit is on SEO and GEO. Search buying is on paid search. Social beside it is on paid social. Rates are on pricing. The door is contact. The rest of the desk is on the news stream.
Frequently asked questions#
Where does the robots meta tag sit on the 21 October page?
In a row with data-nosnippet and the X-Robots-Tag. It is not the only method. The page lists seven.
Does a noindex tag work if robots.txt blocks the page?
The noindex page says no. The crawler has to reach the page to see the tag. A blocked page can still be indexed from other links.
Is data nosnippet the same as removing a URL?
No. It keeps a passage out of a snippet. Removal of a URL is a different row: change the page, password-protect it, or use noindex.
What is the x-robots-tag for?
The HTTP header form of the same family of rules, useful when one rule should cover a file type. The specification page, updated 24 March 2026, is the grammar.
Do ai overviews go away if you use the AI control?
The help page says an excluded site is not used as an input for AI Overviews, AI Mode, and Discover generative features. Other sites can still appear. Training is a separate token, Google-Extended.
Read more on this topic#
Brief the row, not a ranking myth
If an SEO note needs this controls page read without a fake lift, that is the job.
Want folkfox in your Google results and AI answers? Set folkfox as a preferred source.