A good title tag for an AI-assisted article is a short, unique, accurate description of that one page, written in the page's language, with the site name at most once and no repeated keywords. The H1 should say the same thing in a form that reads well on the page, and the H2s and H3s should form a clean outline a reader could scan without the body text. Google uses the HTML title element for its result headline most of the time, but it rewrites titles that are empty, stale, stuffed or out of step with the page, so the safest title is the one that matches what the page actually delivers. None of these rules changes because a model wrote the first draft; what changes is how carefully you have to check the output.

AI title tags and headings at a glance

  • Title tag: the text inside the HTML title element. It labels the browser tab, bookmarks and history, and is Google's first source for the clickable headline in results, which Google calls the title link.
  • H1: the main visible heading on the page. Google reads it too, and may use it when the title tag looks wrong.
  • H2 to H6: the outline of the article. Screen reader users jump between them, and search systems use them to understand the page.
  • No fixed character limit. Google says there is no limit on the length of a title element; long ones are truncated "as needed, typically to fit the device width" (checked October 2026).
  • AI is not the problem in itself. Google's ranking systems reward helpful content "however it is produced", but mass-producing pages to manipulate rankings breaks its spam policies.

How Google turns your title tag into a title link

Google has used text other than the title element since 2012. In August 2021 it introduced a new system that leans more on text visitors can actually see, such as the main headline and other large, prominent text. In the August 2021 announcement, Google said the title tag was still used more than 80% of the time, and that titles would generally no longer change according to the query. After feedback from site owners it adjusted the system, and its September 2021 follow-up put the figure at around 87%.

Google's current documentation lists the sources it draws on: the title element, the main visual title, heading elements such as the H1, the og:title meta tag, other large and prominent text, anchor text on the page, the text of links pointing to the page, and WebSite structured data. Any change you make needs a recrawl before it shows, which Google says can take from a few days to a few weeks.

The same page names the patterns that trigger a rewrite. These map closely onto the faults AI drafts and templates produce:

  • Half-empty titles, such as | Site Name, where a template field never got filled.
  • Obsolete titles, where the title says 2020 but the visible heading says 2021.
  • Inaccurate titles that promise products or topics the page does not show.
  • Micro-boilerplate, where a group of pages share one title and only the content differs.
  • No clear main title, where several headings carry the same visual weight. Google may then use the first one.
  • A language or script mismatch between the title and the page's main text.

Google also handles the site name separately. Results show a site name for the whole domain above the per-page title link, and Google may drop a site name from the title link if it repeats the one already shown.

How to write a title tag for an AI-assisted article

Describe the page, not the topic

Google's guidance asks for "descriptive and concise" titles and warns against vague labels such as "Home" or "Profile". The GOV.UK content guidance on writing clear titles gives a useful test: a title should make sense on its own, in search results and in a list of documents. Its example contrasts "Hazardous waste – new process" with "How to dispose of hazardous waste in your area".

Use the words searchers use

GOV.UK tells its editors to find out what the public calls a subject by looking at common search terms, and to put those terms first, because the official or internal name may not be what people type. Google's SEO starter guide makes the same point with "charcuterie" and "cheese board", while adding that its language matching systems can connect a page to queries that use different words.

Keep it short enough to survive truncation

Neither Google nor the HTML standard sets a hard length. The common "60 characters" figure is a rule of thumb, not a Google rule. GOV.UK sets its own house limit of 65 characters including spaces, because it finds that Google cuts titles off at around that point and longer titles are harder to understand; it allows exceptions when extra length is essential. A practical aim is to put the words that identify the page in the first 50 or so characters, so a cut does not remove them.

Brand once, with a delimiter

Google suggests putting the site name at the start or end of each title, separated by "a hyphen, colon, or pipe", and keeping longer brand descriptions for the home page.

Say it once

Google's own example of keyword stuffing is "Foobar, foo bar, foobars, foo bars". Its starter guide states plainly that keyword stuffing is against its spam policies. One clear phrase for the main subject is enough.

Getting H1, H2 and H3 headings right

The HTML standard says a title should identify the document "even when they are used out of context", in history, bookmarks or search results, and notes that the title "is often different from its first heading, since the first heading does not have to stand alone". That is the real relationship between the two. They need not be identical, and they should not contradict each other. Google's advice on title links suggests making the main title the most prominent text on the page, for example by putting it in the first visible H1.

For the outline below the H1, accessibility guidance is stricter than search guidance:

  • MDN's reference on heading elements advises one H1 per page and no skipped levels. Multiple H1s are allowed by the standard, but MDN does not consider them good practice.
  • The W3C headings tutorial says an H2 should not be followed directly by an H4, though skipping back up a level when a subsection closes is fine.
  • WCAG success criterion 2.4.6, Headings and Labels, asks for headings that describe topic or purpose, and criterion 2.4.2, Page Titled, requires every page to have a descriptive title.

Google takes a looser line on ranking. Its starter guide says headings in semantic order are "fantastic for screen readers", but from a search perspective "it doesn't matter if you're using them out of order", and there is no ideal number of headings. So the case for a tidy hierarchy is mainly about people.

Those people are numerous. In WebAIM's Screen Reader User Survey #10, run in December 2023 and January 2024 with 1,539 valid responses, 71.6% said that navigating by headings is the first thing they do to find information on a long page, and 88.8% found heading levels very or somewhat useful.

GOV.UK's guidance on clear structure adds a writer's checklist: headings should be descriptive (no generic "Introduction"), frontloaded, active where possible, and removable, meaning the text still makes sense without them. GOV.UK also avoids headings phrased as questions, on the grounds that they are hard to frontload and users want answers. Many other publishers use question headings in FAQ sections, so treat that as a house style choice.

Where AI drafts usually go wrong

Language models tend to produce titles that sound like every other title on the subject. Five faults come up repeatedly when editing AI output:

  1. Generic labels. "Introduction", "Benefits" and "Conclusion" tell neither a reader nor a search system what a section contains.
  2. Promises the text does not keep. A model asked for "7 examples" in the title may write five. Google's helpful content questions ask whether the heading or title gives "a descriptive, helpful summary" and whether it avoids "exaggerating or being shocking in nature".
  3. Near-duplicate titles across a batch. When one prompt produces many articles, titles converge. That is exactly the boilerplate pattern Google rewrites, and it makes your pages hard to tell apart in results.
  4. Stale years. A title that says "2025" on a page updated in 2026 is Google's "obsolete title" case.
  5. Template gaps. An empty variable in a title template gives you the half-empty | Site Name title.

Scale raises the stakes. Google's spam policies define scaled content abuse as generating many pages "for the primary purpose of manipulating search rankings and not helping users", and name "using generative AI tools or other similar tools to generate many pages without adding value for users" as an example. Google's 2023 statement on AI-generated content, posted by Danny Sullivan and Chris Nelson, puts the other side: "Using AI doesn't give content any special gains. It's just content."

A sensible workflow is to let the model propose ten or so title and H1 pairs, discard any that overstate the article, then edit the best one by hand against the checks below.

Google's AI headline experiments in 2026

Your title is a request, not a guarantee. In March 2026 Google confirmed to The Verge that it was running a "small and narrow" test of AI-generated headline rewrites in Search, according to Search Engine Journal's report of 22 March 2026. Google said the aim was to "identify content on a page that would be a useful and relevant title to a users' query", that the test was not limited to news sites, and that any broader launch might not use generative AI. One example cut "I used the 'cheat on everything' AI tool and it didn't help me cheat on anything" to "'Cheat on everything' AI tool". The same report describes a similar experiment in Discover, which Google called a small UI test in December 2025 and treated as a feature by January 2026.

Google's title link documentation offers no opt-out from rewrites, and results do not say when a title has been changed. Check the current status before relying on any of this (checked October 2026). In practice the defence is the same as before: a title that already matches the page and the query leaves a rewriting system little to improve.

What to check before you publish

  • The title is unique on your site and makes sense in a search result with no other context.
  • The words that identify the page sit in roughly the first 50 characters.
  • No word or phrase is repeated for keyword purposes; the brand appears once, after a hyphen, colon or pipe.
  • Any year, number or count in the title matches the body.
  • The title and the H1 describe the same thing, and the H1 is the most prominent text on the page.
  • There is one H1, and the H2s and H3s nest without skipped levels.
  • No generic section headings are left over from the AI draft.
  • The og:title from the Open Graph protocol matches or closely tracks the title, since Google lists it as a title source and social platforms use it for link previews.
  • After publishing, the Search Console Performance report shows which pages have the highest and lowest click-through rates, so you can spot titles that are shown but not chosen. Search the exact page now and then to see what Google displays.

Frequently asked questions

Should the title tag and the H1 be identical?

They do not have to be. The HTML standard expects the title to stand alone out of context and notes that it often differs from the first heading. What matters is that both describe the same page, so Google has no reason to pick some other text.

Is there a maximum title tag length?

Google sets none, but it truncates titles to fit the screen. GOV.UK uses a 65-character house limit; many publishers aim for 50 to 60.

Will Google penalise a title because AI wrote it?

Google says it rewards quality "however it is produced". Its spam policies target mass-produced pages that add no value, along with keyword stuffing.

Do multiple H1s hurt rankings?

Google says heading order does not matter for search, but MDN and W3C guidance advise a single H1 and no skipped levels for accessibility.