Should You Write Titles for AI Answers or for Clicks?
Should you write titles for AI answers or for clicks?
As of September 2026, write for clicks and let the AI answer take care of itself. The title element is a weak lever on whether a model cites you and a strong lever on whether a human chooses you. Optimising the strong lever for the weak outcome is a bad trade, and most of the advice telling you otherwise is guessing.
This question comes up in almost every content review we run now. A team has read that AI answer engines summarise pages, and they want to know whether their titles should change shape to suit a machine reader. It is a reasonable question with an unsatisfying answer, and the unsatisfying part is where the useful detail lives.
Here is what is actually known, what is not, and how we would write a title today.
How much are clicks actually worth now?
Less than they were, and that is the honest starting point. Pew Research Center analysed the browsing data of 900 US adults through KnowledgePanel Digital, covering 68,879 unique Google searches from March 2025, of which 12,593 produced an AI summary. Users who encountered an AI summary clicked a traditional search result in 8 percent of visits. Those who did not encounter one clicked nearly twice as often, in 15 percent of visits.
The in-summary link fared worse still. Clicking a link inside the AI summary itself happened in just 1 percent of all visits to pages with such a summary.
Set that against scale. Google's own announcement for website owners, updated August 31, 2026, says AI Overviews has over 2.5 billion monthly active users and AI Mode has surpassed one billion monthly users. So a shrinking click rate is being applied to an enormous and growing surface, which is why the total picture is harder to read than the percentages suggest.
Does your title even reach the AI answer?
Not in the way people assume. When a model generates an answer, it is working from page content, not from your title element as a special field. The title is one signal among many about what the page is, and there is no published mechanism by which a cleverly worded title increases the chance of citation.
We want to be careful here, because this is exactly the kind of gap where confident advice appears without evidence. Nobody outside these companies knows the retrieval and ranking internals of AI Overviews, AI Mode, ChatGPT search or Perplexity. Anyone telling you the optimal title format for AI citation is describing a hypothesis.
What is documented is how Google builds the title link you see in search results, and that is worth knowing precisely.
What does Google actually use to build a title link?
Nine things, and Google lists them explicitly. Content in the title element, the main visual title shown on the page, heading elements such as h1, content in og:title meta tags, other content that is large and prominent through style treatments, other text contained in the page, anchor text on the page, text within links that point to the page, and WebSite structured data.
That list should change how you think about the job. Your title element is the first of nine inputs, not the only one. If your title says one thing and your visible h1 says another, you have given Google a choice, and Google will sometimes take the other option.
Google is also clear that it does not always use what you wrote, and that it has to recrawl and reprocess a page to notice updates. If you changed a title last week and the result looks unchanged, that is expected behaviour rather than a penalty.
How long should a title be?
Long enough to be specific, short enough to survive truncation. Google's documentation says there is no limit on how long a title element can be, and that the title link is truncated in search results as needed, typically to fit the device width.
Notice what that does and does not say. There is no character limit and no penalty for length. There is a display constraint that varies by device, which means any specific number you have been given is a convention rather than a rule.
Our practical approach is to put the distinguishing words first and treat everything after roughly sixty characters as a bonus that some readers will see. That is not a Google rule. It is what happens when you accept that truncation is a display fact you cannot control.
What is the real difference between the two targets?
They pull in opposite directions on exactly one dimension, which is curiosity. A title that earns a click often withholds something. A title that serves a machine summariser should withhold nothing.
| Dimension | Writing for clicks | Writing for AI answers |
|---|---|---|
| Specificity | High | High |
| Plain language | High | High |
| Curiosity gap | Sometimes useful | Actively unhelpful |
| Brand suffix | Useful for trust | Neutral |
| Matching the page content | Essential | Essential |
Look at how much of that table agrees. Four of the five rows want the same thing. The entire conflict is one row, and it only applies to the subset of titles that were going to use a curiosity gap in the first place. For most B2B pages, which are trying to say plainly what something is, there is no trade-off at all.
So when does the conflict actually matter?
On opinion and analysis pieces, where a provocative title is a legitimate tool. A headline that poses a question without answering it will earn clicks from humans and will give a summariser nothing to extract.
The resolution we use is to let the title be interesting and make the first paragraph do the answering. Nothing stops a page from having a curious title and an opening that states the conclusion in the first thirty words. That combination serves both readers, and it costs nothing.
What we would not do is flatten every title into a declarative statement because a machine might read it. You would be trading a measurable human behaviour for a speculative machine one.
What about the pages nobody will ever click?
Some pages exist to be quoted, not visited. Glossary entries, definition pages, specification tables. On those, the click was never realistic, and writing the title as a plain description of the thing is obviously right.
The mistake is applying that logic sitewide. A pricing page, a case study and a service page all still depend on somebody arriving. Treating them like reference material because AI answers exist gives up the visit you actually needed.
The useful split is by page intent. Reference pages optimise for being quoted. Decision pages optimise for being visited. We went further into this territory in building a strategy for zero-click search.
What should you actually change on your site?
Three things, none of them exotic. Make your title element and your visible h1 say the same thing, so Google is not choosing between them. Put the distinguishing words at the front. Remove boilerplate repeated across pages, which Google explicitly names as something to avoid alongside keyword stuffing.
Then brand concisely at the end, which Google's own best practices endorse, using a delimiter such as a hyphen or a colon. Brand at the end is close to free: it builds recognition for readers who scan many results and it costs you nothing in the part of the title that gets read first.
If you want the mechanics in more depth, we covered them in writing title tags and meta descriptions that hold up, and the underlying guidance has not changed.
How should teams think about this going forward?
Optimise for the thing you can measure, and hedge cheaply on the thing you cannot. Clicks are measurable in Search Console. AI citation is partially visible at best. When one outcome is observable and the other is a hypothesis, put your effort behind the observable one and make sure you are not actively hurting the other.
The cheap hedge is exactly what good writing already does: say clearly what the page is, answer the question early, keep the page and the title consistent. None of that costs you a click, and all of it is what every published guideline about being useful to machines has said. We wrote about the broader version in how to get cited by AI search engines.
The trap to avoid is rewriting hundreds of titles on the strength of advice nobody can substantiate. That is a large amount of work bought with a guess, and the click data is one of the few things in search right now that is not a guess.
If you want a straight read on whether your titles are earning what they should, or how your pages are showing up in AI answers, we are happy to walk through it. You can find us at phoenix.studio.
Want a site that performs like this?
Tell us about your project. We will come back with a clear next step, no pressure.
This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.
Have a project like this?
Tell us where you want to go. We'll tell you how we'd get you there.