A title and a meta description were written for one surface: a list of blue links where a person reads two lines and decides whether to click. On an answer surface the person often never sees either. The assistant reads the page, lifts a passage, and shows a short source link beside its answer. So the question for an SEO is what these two elements still do once the content is being extracted rather than browsed.
The short answer: the title still matters, the meta description matters less than it did, and neither one is where the answer lives. The title is the cheapest label a system has for what a page is about, and Google builds the visible link from it and from the page’s headings. The meta description is a suggestion Google says it uses only sometimes, and nothing documented says an assistant reads it at all. Write the title to name the answer topic, write the description as a plain summary, and put the actual answer in the body.
This post covers what Google documents about title links and snippets, what is undocumented on the assistant side, how to write both elements for extraction, and how to check them across a site. It sits under Answer Engine Optimization: a page-level guide.
What does Google say a title link is built from?
Google builds the title link from several sources, and the <title> element is only the first of them. Its title link documentation lists the inputs: the content of <title>, the main visual title on the page, heading elements such as <h1>, og:title meta tags, other large and prominent text, other text on the page, anchor text on the page, text in links that point to the page, and WebSite structured data.
The consequence: a title that disagrees with the H1 gives Google a reason to choose something else. If the <title> says “Pricing | Acme” and the H1 says “Plans for teams of every size,” the link shown may be neither. An extractor needs the same consistency: one topic, stated the same way in the title, the H1, and the first paragraph.
Google also states that there is no limit on how long a <title> element can be. The title link is truncated as needed, typically to fit the device width. Length windows you see in audit tools, including GEOBee’s, are house conventions for display, not rules Google publishes. Treat them as a warning about truncation, not a ranking threshold.
Is the meta description still used?
Sometimes, and Google says so directly. Its snippet documentation states that snippets are primarily created from the page content itself, and that Google sometimes uses the meta description if it might give users a more accurate description than other parts of the page. Like the title, there is no limit on meta description length; the snippet is truncated to fit the device.
So in classic results, the body text already competes with your meta description for the snippet. A meta description that summarizes the page honestly gets picked more often than one stuffed with keywords. Google calls out that “long strings of keywords” in a description do not give users a clear idea of the page.
The snippet controls belong to the same page. nosnippet, max-snippet, and the data-nosnippet attribute limit what Google can show from a page. That matters more on answer surfaces than the description does, and nosnippet, max-snippet and data-nosnippet on AI surfaces covers those directives on their own.
What changes on an answer surface?
What changes is that the body, not the head, supplies the text. Google’s AI features documentation states that to be shown as a supporting link in AI Overviews or AI Mode, a page must be indexed and eligible to be shown in Google Search with a snippet, and that there are no additional technical requirements. It does not mention titles or meta descriptions. AI Overviews eligibility: every technical requirement walks through that list.
What is documented, then, is eligibility. What is not documented is how the title and description are weighted when an AI feature picks a supporting link, or whether they are read at all. The same is true outside Google: the crawler documentation from OpenAI, Anthropic, and Perplexity describes which agents fetch pages and how to control them, not which page elements their answer systems read. Anyone who tells you the exact role of a meta description in an assistant’s answer is guessing.
Here is what can be said without guessing:
| Element | Classic results | Answer surfaces |
|---|---|---|
<title> | One of several inputs to the title link Google shows | Undocumented; a likely label for the page, since it is the most compact statement of its topic |
| Meta description | Used sometimes as the snippet; body text often wins | Undocumented; the answer text is lifted from the body |
| H1 | Input to the title link | Part of the page’s main content, read with the body |
| Body text | Primary source of snippets | Source of the passage that gets lifted |
nosnippet / max-snippet | Limits the snippet | Limits eligibility as a supporting link in Google’s AI features |
The title and description are labels; the answer is in the body, which is why what an answer engine reads focuses on the main content block rather than the head.
How should you write a title for extraction?
Name the topic the page answers, in the words a person would use to ask about it, and put the specific thing first. A title is the shortest description of the page any system will see, so it should survive being read without the page around it.
Four rules hold up:
- Name the subject, not the category. “Pricing | Acme” labels a section. “Acme pricing: plans, seat limits, and annual billing” names what the page answers.
- Match the H1 in substance. They do not need to be identical, but they should describe the same page. Google lists both as inputs; a mismatch invites a rewrite.
- Put the distinguishing words early. Truncation happens at the end. A site name, a separator, and a category before the topic push the topic off the visible line.
- One page, one title. Duplicate titles across URLs tell every system the pages are interchangeable. Google’s guidance says the same thing about boilerplate titles repeated across a site.
A title can be phrased as a question, but it need not be; matching questions is the job of the headings, covered in question headings vs. keyword headings.
Here is an illustrative before and after, for a fictional support page:
| Title | |
|---|---|
| Before | Support Article 4471 – Help Center – Acme Robotics |
| After | Resetting the ARM-X9 controller after a firmware update – Acme Robotics |
The second names the product, the task, and the condition, before the site name.
How should you write a meta description now?
Write it as a two-sentence summary of what the page says, including the main answer if one fits. Even when it is not shown, it is a statement of the page’s subject that agrees with the body. When it is shown, it is a summary a person can act on.
Three things to avoid:
- A teaser that withholds the answer. “Find out how to reset your controller” describes a page; “Hold the reset button for ten seconds after the update finishes, then re-pair the pendant” summarizes it. The second is useful wherever it appears.
- A description that disagrees with the body. If the meta says “free shipping on all orders” and the page says “free shipping over a threshold,” you have published two answers. Google may show either.
- The same description on many pages. Google recommends unique descriptions per page and notes that large sites can generate them programmatically from page data. A template that fills in the product name and two attributes beats one sentence repeated on every product URL.
Do not spend effort here before the body is right; the description cannot fix a page whose first paragraph does not answer its heading. The 40–60 word answer block is the higher-return edit; the meta description is often a lightly trimmed copy of it.
How do you check one page by hand?
Open the page, view the source, and read four things side by side: the <title>, the <meta name="description">, the H1, and the first paragraph of the main content. Then ask three questions.
- Do all four describe the same topic?
- Does the title name the specific subject before any site name or category?
- Does the description summarize what the body actually says, including the answer if one fits?
If any answer is no, the fix is an edit. Then search for the page in Google and compare the title link and snippet it shows with what you wrote. A rewrite usually came from the H1 or body text, which tells you which element Google found more descriptive.
How do you check them across a site?
Use a crawler export and a spreadsheet; no terminal is needed. Any crawler that records the title and meta description per URL can export them. Paste the export into a sheet and add three columns:
| Column | Formula | What it catches |
|---|---|---|
| Title length | =LEN(B2) | Titles likely to truncate, or empty ones |
| Duplicate title | =COUNTIF(B:B, B2)>1 | Templates that stamp the same title on many pages |
| Title starts with site name | =LEFT(B2, 5)="Acme " (adjust to your brand) | Titles that put the label before the topic |
Repeat the first two for descriptions. Sort by the duplicate column first: a duplicate title is almost always a template fault, and one template fix clears hundreds of rows.
For the topic-match question, filter to your highest-value templates and read the title and H1 columns side by side. That part is judgment.
How GEOBee scores titles and descriptions
GEOBee checks titles and meta descriptions in two lenses. In the AEO lens, Title & meta is worth 10 of 100 points and checks that both fall within healthy length windows. In the SEO lens, Title is worth 15 and Meta description 10. The three lenses lists every weight.
The length windows default to 30–60 characters for titles and 50–160 for descriptions, and you can change them in the Scoring Rules tab described in Crawl settings. As noted above, those are display conventions, not limits Google publishes, so set them to your own house style. Duplicate titles and meta descriptions across URLs are raised as site-wide findings rather than per-page ones.
What GEOBee cannot score is whether a title names the right topic. That is still a read. The score tells you which pages are missing or truncating a label; it does not tell you, and does not claim, that a better title will get a page cited.
If you want the title and description columns for every URL in one export, download GEOBee and run a quick scan during the 7-day trial.