Before any of the usual advice about character counts and power words, one fact reframes the whole exercise: most of what you write here never appears in the search results. Google replaces it.
The number that changes the job
A 2025 study covering thousands of keywords across commercial, informational and sensitive categories found Google rewrote roughly 76% of title tags — up from about 61% in a 2023 study using similar methods. The trend is upward.
Meta descriptions fare worse. Research puts replacement at around 68% on desktop and 71% on mobile. Roughly seven times in ten, the description a searcher reads is something Google assembled from your page, not the sentence you carefully wrote.
Google's reasoning is defensible. Your description is written once. The page might rank for hundreds of different queries. When your sentence doesn't match the specific thing someone typed, Google pulls text from the page that does.
The reframe this article runs on You are not writing what gets displayed. You are writing the strongest available input to something Google will probably rewrite — so the goal is to influence the rewrite, not to control the display.
That isn't an argument for skipping the work. The descriptions Google keeps are disproportionately the well-matched ones, and the title you write remains the single biggest influence on whatever ends up shown. But it does change what "good" means, and most advice on this topic is still written as though you had full control.
How to reduce your odds of being rewritten
This is the section missing from nearly every guide, and it's the practical core.
Rewrites cluster around identifiable causes. Fix these and your survival rate goes up measurably.
| Trigger | Why it prompts a rewrite | Fix |
|---|---|---|
| Too long | Truncation makes the title unusable as displayed | Keep the meaningful part inside the visible width |
| Keyword stuffing or repetition | Reads as written for machines, not readers | State the topic once, clearly |
| Title doesn't match the H1 | Google prefers the visible heading when they conflict | Keep them closely aligned, not necessarily identical |
| Boilerplate branding crowding it out | Site name repeated on every page adds nothing | Brand at the end, or omit on long titles |
| Doesn't match the query it ranks for | The most common cause of all | Check which queries actually drive impressions |
| Generic description | Nothing in it is specific enough to serve any query | Name the specific thing the page delivers |
The fifth row deserves emphasis because it's the one people can't see without looking. The queries your page actually ranks for are frequently not the ones you wrote it for. Open Search Console, filter to the page, and read the top twenty queries by impressions. If your title doesn't reflect any of them, you've found your rewrite cause — and probably a bigger opportunity than the title itself.
Length: the advice everyone gives is slightly wrong
"Keep titles under 60 characters, descriptions between 150 and 160." Fine as a rule of thumb, and not actually how truncation works.
Google truncates on pixel width, not character count. A title of wide capital letters runs out of room considerably sooner than one of the same length in narrow lowercase characters. Which means two titles at exactly 58 characters can behave completely differently.
Mobile is stricter again. Displayed descriptions on mobile often land nearer 105–120 characters, and shorter still when a publication date is shown. Since mobile is the majority of impressions for most sites, that's the constraint that matters.
The workable rule: front-load the meaningful content into the first 100 characters of a description and the first 50 of a title. Anything after that is a bonus if it survives. Write so the sentence still makes sense if it's cut off mid-way — because on some devices, for some queries, it will be.
SE Ranking
Fast-growing affordable SEO platform perfect for agencies and small businesses
Best for: Budget SEO Platform
What actually earns the click
Assuming your title survives, or the rewrite still resembles your intent, what makes someone choose your result?
Specificity beats persuasion. "Complete Guide to X" is a category. "X in 12 Steps, With the 3 That Usually Fail" is a page. Numbers, named constraints and concrete scope all outperform adjectives, because they let someone predict what they'll get.
Match the stage of the question. Someone searching "what is X" wants orientation; someone searching "X vs Y" wants a decision; someone searching "buy X" wants a transaction. A title written for the wrong stage loses the click even when the page would have served them — which is why keyword research and snippet writing are the same job done at two moments.
Say what's different about your version. Every result on the page claims to answer the question. The one that says how it answers differently — original data, a specific year, a named method, a stated point of view — gives someone a reason to prefer it.
Front the answer in the description. Rather than teasing ("Discover the secrets to…"), state the substance. Withholding to force a click is a strategy from an era when the click was the only way to get the information, and that era is ending.
The 2026 complication: your snippet has company
An assumption underpinning all traditional snippet advice has quietly broken.
That advice assumed your result competes against nine other blue links. Increasingly it competes against an AI-generated answer sitting above the whole list — one that may already have answered the question, sometimes using your page as a source. Click-through on queries carrying AI answers has fallen sharply, and the gap between ranking and receiving a visit has widened accordingly. That's the picture set out in how AI Overviews have reshaped SEO and in our mid-year state of SEO.
Two practical consequences for how you write snippets.
The snippet's job shifted from "answer the question" to "offer what the answer didn't." If a summary above you has already covered the basics, a title promising the basics is redundant. A title promising the specific, the current, the tested, or the contrary is not.
Clarity now serves two readers. The plainly-stated, self-contained sentence that makes a good description is also the kind of text that gets extracted into AI answers — the overlap between snippet craft and generative engine optimisation, and the same property discussed in content marketing in the age of AI answers. Writing for extraction and writing for clicks have converged more than they've diverged.
Testing, which almost nobody does
Snippet advice is delivered with total confidence and almost never with a measurement method. Here's one.
- Pick a page with real traffic. Under roughly a hundred clicks per period won't produce a readable result. Start with pages ranking in positions four to ten — high impressions, unrealised clicks.
- Record the baseline. Four weeks of impressions, clicks, average position and CTR from Search Console.
- Change one thing. Title or description, not both. Otherwise you learn nothing about which one moved the number.
- Wait four weeks. Shorter periods pick up seasonality and noise.
- Compare — and check position first. This is the step people skip. If average position moved, your CTR comparison is contaminated: CTR rises naturally as position improves. Only if position held roughly steady does a CTR change tell you anything about the snippet.
That fifth point invalidates a large share of the title-testing case studies circulating online, which report CTR gains without controlling for position movement. Treat any "we changed a title and got +40% clicks" claim with the same scepticism you'd apply to any uncontrolled before-and-after — a caution that applies to most marketing measurement right now.
Where to start, and it isn't everywhere
Rewriting every title on a site is a poor use of a week. The returns concentrate in three places:
- High impressions, low CTR, positions 4–10. People are seeing you and choosing someone else. Best available return, and Search Console will list these in a couple of clicks.
- Commercially important pages, whatever their position. A small CTR gain on a page that converts beats a large one on a page that doesn't.
- Pages ranking for queries you didn't target. These are usually mismatched titles, and fixing the match often lifts both CTR and relevance.
Two cautions before you start editing. Titles carry some direct relevance weight, so changing them on pages that rank well is a real risk — if a page performs, leave it alone. And treat title changes on high performers as you would any change to a working system: one at a time, with the ability to roll back.
Descriptions carry no such risk. They aren't a ranking factor, which makes them the safer place to experiment and the sensible first move if you're nervous.
What to write, by page type
Patterns worth adapting rather than templates to copy — remember that anything used verbatim across a category becomes boilerplate and invites exactly the rewriting you're trying to avoid.
Guides and how-tos: the specific outcome plus the constraint that makes it credible. Not "Guide to X" but "How to Do X Without Y." The constraint is what signals you've actually done it.
Comparisons: name both things, and state what the comparison is on. Someone searching a comparison has already narrowed their options; they want a basis for deciding, not an overview of both.
Product and category pages: what it is, one differentiator, and the practical fact that removes hesitation — availability, range, price position. This is the same reasoning that governs product page optimisation, applied a step earlier in the journey.
Local pages: service plus place, in natural language. Location stuffing reads badly and rewrites frequently.
Homepages: what the business does, for whom. The most commonly wasted title on any site — brand name plus a tagline nobody outside the company understands is the default, and it tells a searcher nothing.
That last one is worth checking today. If your homepage title doesn't say what you do, it's costing you clicks from people who've heard of you and aren't sure you're relevant — one of several quiet ways a site loses customers without anyone noticing.
The short version
Google rewrites roughly three-quarters of titles and seven in ten descriptions, so you're writing an input rather than an output. Reduce your rewrite odds by keeping titles concise, aligned with the H1, free of stuffing, and matched to the queries the page actually ranks for. Front-load the first 100 characters, because pixel width and mobile truncation cut earlier than the character counts suggest. Write specificity rather than persuasion, and offer what the AI answer above you didn't. Then test properly — one change, four weeks, and check whether position moved before you believe the CTR.
Ranking well and still not getting the clicks?
We audit search visibility end to end and fix the pages losing traffic they've already earned.
Explore SEO Services →Frequently asked questions
Does Google actually use the title tags and meta descriptions you write?
Often not. A 2025 study of thousands of keywords found Google rewrote roughly 76% of title tags, up from about 61% in a 2023 study. Meta descriptions are replaced even more frequently, with research putting the rate near 68% on desktop and 71% on mobile. That does not make writing them pointless — the ones Google keeps tend to be the well-matched ones, and what you write is the strongest input to whatever gets generated. But the realistic goal is influencing a rewrite rather than controlling a display.
How long should a title tag and meta description be?
Titles are safest under about 60 characters and descriptions between roughly 150 and 160, but the real constraint is pixel width rather than character count, so a title full of wide capitals truncates sooner than one of the same length in narrower characters. Mobile truncates considerably earlier — displayed descriptions often land near 105 to 120 characters. The practical rule is to front-load the meaningful part into the first 100 characters so the message survives whichever cut-off applies.
How do you stop Google rewriting your title tag?
You cannot prevent it, but you can reduce the odds. Rewrites are more likely when the title is too long and gets truncated, when it is keyword-stuffed or repetitive, when it does not match the visible H1 or page content, when boilerplate branding crowds out the substance, or when the title poorly matches the query the page ranks for. Titles that are concise, accurate, consistent with the page heading, and aligned with real search intent survive at noticeably higher rates than titles written to satisfy a keyword checklist.
Are meta descriptions a ranking factor?
No. Google has stated for years that meta descriptions are not used as a ranking signal. Their value is indirect: a description that accurately reflects the page and matches search intent can improve click-through rate, and behaviour on the results page can influence how a page performs over time. Title tags are different — they do carry some direct relevance weight — which is why title changes should be made more carefully than description changes.
How do you test whether a new title improves clicks?
Use Search Console rather than guessing. Record impressions, clicks, average position and CTR for the page over the four weeks before the change, make one change at a time, wait at least four weeks, then compare like for like. The critical check is average position: if ranking moved, the CTR comparison is contaminated and tells you nothing about the title. Pages with under roughly a hundred clicks per period will not produce a readable result, so test on pages that already have traffic.
