Headline Analyzer
Check a headline against four measurable properties — word and character count, power-word count, digits, and the share of words outside a common-word list — and get a transparent 0-100 checklist score.
Headline Analyzer
What this score is. It is RECATOOLS' own checklist, not a click-through prediction. The tool measures four things it can actually count — word and character length, how many words appear in the power-word lists below, how many digits the headline contains, and what share of its words fall outside a 150-word common-word list — and adds the points shown in the FAQ. No published study sets those weights and no click data is behind them. A high score means the headline is specific and scannable; whether it earns a click is a question only your own testing answers.
How to use the Headline Analyzer
Type or paste your headline
Enter the headline candidate — blog title, article title, newsletter subject line, ad headline. Tool scores in real-time. Test multiple variants quickly by editing in place.
Read the 0-100 score
Four components, all of them counting rules you can check by hand: length (25 points for 6-14 words, plus 5 for 40-65 characters), power words (25 for two to four matches in the lists), numbers (15 for any digit) and uncommon ratio (20 for 40-70% of words outside the common-word list), with 5-point bonuses for a question mark and for opening on how, why, what, when, where or who. The bands are ours: 80+ reads as specific and scannable, under 40 as generic. They are not calibrated against click data.
Iterate based on the advisory
The advisory names the component that is costing you points — too short, no power word, no number, too many common words. Change one thing at a time and watch the score move, so you can see which rule fired. Be aware that you can also game it: pad a vague headline with a digit and two power words and it will score well while saying nothing.
Test 5 headline variants
Write several candidates, score each, then shortlist on your own judgement rather than on the number. The score cannot tell you whether a headline is honest about the article, whether the topic matters to your readers, or whether the phrasing lands in your market. Only a live test measures click-through, and this tool does not attempt to predict it.
Headlines — the single highest-leverage line you'll write
The argument that a headline deserves disproportionate effort is old and comes from advertising rather than from the web: David Ogilvy's Confessions of an Advertising Man (1963) reports that roughly five times as many people read the headline as read the body, and concludes that most of a copywriter's budget is spent by the time the headline is finished. Web research reaches a compatible conclusion by a different route. Nielsen Norman Group's work on microcontent (Loranger and Nielsen, 29 January 2017) finds that people scan rather than read, and advises moving the words that carry meaning to the front of a title, cutting non-essential words, and dropping leading articles — because the opening words are the part that survives a scan.
What this tool actually measures
Four counts, and nothing more. Length, in words and characters. Power words: how many of the headline's words appear in five lists shipped with the tool — emotional, authority, curiosity, practical and urgency — which were assembled by us, not drawn from a published lexicon. Numbers: whether the headline contains any digit. Uncommon ratio: the share of words that do not appear in a 150-entry list of English function and filler words, used as a rough proxy for specificity. The points attached to each are set out in the FAQ, and they are our editorial judgement. No click-through dataset informs them, and this page makes no claim about what any of them does to a click rate.
A score is a checklist, not a forecast. It can tell you a headline is vague; it cannot tell you anyone wants to read the article.
What the search engine does with your title
If the headline is also the page title, one thing is documented rather than folklore. Google states plainly that there is no limit on how long a <title> element can be, and that the title link shown in results is truncated as needed, typically to fit the width of the device. So there is no character count that wins and none that loses — a long title is not penalised, it is merely cut. Google also says it may replace your title entirely, drawing on the page's main visual heading, its h1, its og:title, prominent on-page text, or the anchor text of links pointing at it. The 40-65 character band this tool rewards comes from Nielsen Norman Group's scanning guidance, where the point is legibility to a human skimming a list — not a search ranking.
Headline shapes and where each one fits
Four structures worth having in reach, offered as patterns rather than performance claims. A listicle — "N [adjective] [topic] that [benefit]" — suits a roundup where the count is real and the items are genuinely distinct. A how-to — "How to [outcome] without [obstacle]" — suits instructional writing, and sets an expectation the body has to meet step by step. A curiosity headline — "Why [subject] [surprising outcome]" — only works when the article actually resolves the question; when it does not, it is the shape readers recognise as bait. A benefit-led headline — "[Verb] [specific outcome] in [time]" — is the most demanding, because the specific number in it is a promise you have to keep. Pick the shape that matches what you wrote, not the one that scores highest here.
Writing headlines for a second-language audience
This tool's lists are English, and its common-word list is English function words, so its judgement does not carry over to a headline written in Bahasa Indonesia, Malay, Thai, Vietnamese or Filipino — the power-word count will read zero and the uncommon ratio will be meaningless. What does carry over is structural: put the words that carry the topic first, keep the sentence short, and prefer a concrete claim to a clever one. Wordplay, idiom and double meaning ask for cultural fluency that a second-language reader may not have and that a translation will not preserve, which is a reason to save them for body copy where there is room to recover from a miss.
How the 100 points are handed out, and what the score cannot see
Length is worth 30 of the 100 points: 25 for a headline of 6 to 14 words, 15 for 4-5 or 15-16, and 5 otherwise — plus a separate 5 if the character count lands between 40 and 65.
Power words are worth 25, and the curve is deliberately humped: two to four matches score 25, a single match scores 15, and five or more drops you to 12. Stuffing is penalised, not rewarded.
The power-word lists hold about 120 entries across five categories and were written by us. They are not a published lexicon, and a good headline can easily contain none of them.
Numbers are worth a flat 15 for any digit at all. "7 ways" and "2026" score identically, which is a limit of the rule rather than a claim that both are equally useful.
The uncommon ratio is worth 20 and is measured against a 150-word list of English function words — the, is, of, how, new, best. It is a proxy for specificity, and a crude one.
Two 5-point bonuses remain: one for a question mark, one for opening on how, why, what, when, where or who. Both can be earned by a headline that promises nothing.
The score is blind to meaning. It cannot tell whether the headline describes the article, whether the number in it is true, or whether the claim is one you can keep.
Google states there is no limit on the length of a title element; the title link is simply truncated to fit the device. The 40-65 character band here is about human scanning, not ranking.
Google may not show your title at all. It also draws on the page's main visual heading, its h1, its og:title, prominent on-page text and the anchor text of inbound links.
Nielsen Norman Group's microcontent guidance (2017) is the one piece of research behind any of this: move meaning-carrying words to the front, cut non-essential words, drop leading articles.
Frequently Asked Questions
-
Additively, and you can reproduce it by hand. Length: 25 points for 6-14 words, 15 for 4-16, 5 otherwise, plus 5 more if the headline is 40-65 characters. Power words: 25 for two to four matches, 15 for exactly one, 12 for five or more, 0 for none. Numbers: 15 for any digit. Uncommon ratio: 20 for 40-70% of words outside the common-word list, 12 for 30-80%, 5 otherwise. Then 5 for a question mark and 5 for opening on how, why, what, when, where or who, capped at 100. Those weights are ours. They are not derived from click data, and no published study sets them.
-
No, and they are not comparable. CoSchedule scores against their own model and their own word lists, neither of which is published, so a 72 there and a 72 here are two different numbers that happen to share a scale. The one thing this tool offers that a closed scorer cannot is that every rule is written out above, so you can decide whether you agree with it. If two analysers disagree about a headline, that tells you the analysers disagree — not which headline is better.
-
No, and it does not predict click-through at all. The tool counts four properties of the text and adds up points; it has no click data behind it and cannot have any, because it never sees how your headline performs. Topic relevance and audience swamp everything it measures — a specific, well-formed headline about something nobody wants to read will score high and be ignored. Treat a low score as a prompt to look again at a vague headline, and treat a high score as no evidence whatsoever.
-
Only when they are accurate. The words in these lists earn their place by being informative, not by being emphatic: "free" tells the reader about the price, "step-by-step" about the format, "printable" about what they get. Those describe the article. "Amazing", "incredible" and "jaw-dropping" describe the writer's opinion of it, and a headline built from them tells the reader nothing. The scoring caps out at four matches and drops at five precisely because stacking them is a legible tell — but a headline with none of these words and one concrete fact is better than either.
-
The question assumes a ranking this tool cannot supply — it has no performance data for any format. What it can say is which shape fits which article. A listicle needs genuinely separable items and a count that is not padded. A how-to needs the body to actually be a procedure. A question headline needs the article to answer it, and to answer it in a way the reader could not have guessed from the headline; if they could, the question was decorative. Choose by what you wrote.
-
Partly. The counting works on any short line of English, but two of the rules pull the wrong way for email. The length band here rewards 6-14 words, which is long for a subject line that has to survive truncation in a mobile inbox list. And several words that score points here — free, guaranteed, act now, limited, last chance — are exactly the ones spam filters weight against, so the tool will reward you for a subject line that lands in a junk folder. Use it for the word-count and specificity read; ignore the power-word score for email.
-
This tool has no view on it and neither do we, in the absence of data we can point at. The widely repeated claim that negative headlines beat positive ones is usually traced back to prospect theory, which is about how people weigh gains and losses in gambles, not about how they click on articles; the leap is someone else's, not the research's. What is safe to say is narrower: a headline that states a specific failure is often more informative than one that promises a vague improvement, and that is a property of specificity, which this tool does measure.
-
If you have the traffic for it, yes, because a live test is the only thing here that measures anything. Change one element at a time so the result is attributable, and work out the sample size you need before you start rather than stopping when a variant looks ahead — early leads in small samples reverse routinely. Most email platforms include subject-line testing; for pages you will need a testing tool, since Google Optimize was retired in September 2023. If your traffic is too thin for significance, say so honestly and decide on editorial grounds instead of dressing up a coin flip.
-
No. Everything happens in JavaScript on the page you are looking at: the headline is read from the input, counted, and written back into the results box. The tool's code contains no
fetch,XMLHttpRequestor form submission of any kind, and nothing is written to storage either, so reloading loses your work. Open DevTools → Network and type — no request appears. Unpublished headlines are safe to paste here. -
Three on this site do adjacent jobs. The SEO Title Tag Checker looks at the same line as a page title, where the constraint is truncation rather than tone. The Meta Description Checker handles the sentence underneath it in a search result, which should add to the headline rather than repeat it. The Reading Time Calculator is worth a glance when the headline promises a quick read — it is the cheapest way to find out that it does not.
Method & sources
How it computes
A transparent 0-100 checklist of RECATOOLS' own devising — not a click-through model, and not derived from any published dataset. It adds four counts made in the browser: length (25 points for 6-14 words, plus 5 for 40-65 characters), matches against five power-word lists written for this tool (25 for two to four), any digit (15), and the share of words outside a 150-entry common-word list (20 for 40-70%), with 5-point bonuses for a question mark and a how/why/what opening.
What this tool implements
- The weights are editorial judgement and the page says so in four places — the results panel, the how-to, the about section and the scoring FAQ. There is no external authority for them and none is implied.
- Full power-word banding: 25 points for two to four matches, 15 for exactly one, 12 for five or more, 0 for none — the curve falls past four so that stuffing is penalised rather than rewarded.
- The one documented external fact the tool relies on is Google's own statement that a title element has no length limit and that the title link is truncated to fit the device; the 40-65 character band is therefore attributed to human scanning guidance, not to search ranking.
- Nielsen Norman Group's microcontent guidance (Loranger and Nielsen, 2017) supplies the only research-backed advice on the page: move meaning-carrying words to the front, cut non-essential words, drop leading articles.
- The power-word and common-word lists are English-only and the page states that the score is meaningless for a headline written in Bahasa Indonesia, Malay, Thai, Vietnamese or Filipino.
- Everything runs in the browser: the headline is never transmitted and never stored, not even in localStorage.
Sources
- Google Search Central. Control your title links in Google Search results — 'While there's no limit on how long a <title> element can be, the title link is truncated in Google Search results as needed, typically to fit the device width'; also the list of sources Google may use instead of the title element (main visual title, h1, og:title, prominent text, anchor text, WebSite structured data). Page last updated 2025-12-10 UTC; fetched 2026-09-13. https://developers.google.com/search/docs/appearance/title-link
- Loranger H, Nielsen J. Microcontent: A Few Small Words Have a Mega Impact on Business. Nielsen Norman Group; 29 January 2017 — 'Move the keywords to the front the of the title to catch people's attention and to support scanning'; 'Shorter text is easier to read (and scan)'; page titles of 40-60 characters. https://www.nngroup.com/articles/microcontent-how-to-write-headli…
- Ogilvy D. Confessions of an Advertising Man. New York: Atheneum; 1963 — the origin of the claim, repeated on this page with attribution, that about five times as many readers see the headline as read the body copy.
What can make this go out of date
- none at runtime — the word lists ship with the tool and no figure on the page moves on a publication schedule.
Pick up where you left off
Stored only in this browser — never sent to our servers.