Frequently Asked Questions
Face shape, photo analysis, privacy and the attractiveness test.
Short answers to the questions that come up most often. The first three sections cover face shape and the three ways to find yours; the fourth covers the AI attractiveness test; the last two cover styling and the site itself. If your question is not here, the contact page reaches us.
Face shape basics
What the categories mean, and what they do not.
- Is a face shape a scientific fact?
- No. Face shape categories are a styling convention, not a biological fact. Reasonable people disagree about borderline faces, and our own rule set publishes the boundaries it uses so you can see exactly where your face sits relative to them.
- Why seven face shapes and not eight?
- Some styling guides treat "inverted triangle" as its own category and others use it as a synonym for heart. The boundary between the two is not consistent enough to classify against, so we keep it inside heart and say so rather than inventing an eighth prototype.
- What do the seven face shapes actually describe?
- Each label summarises five proportions: face length against cheekbone width, forehead width, jaw width, the angle at the jaw corner, and how much the chin tapers. The label names the nearest reference shape. It is a compact way to talk about proportions, not a separate physical trait.
- How do I find my face shape without guessing?
- There are three independent routes here: a photo analysis that measures 478 landmarks in your browser, a seven-question visual quiz, and a tape-measure calculator that uses four measurements. Each one reports a best match and the margin around it, and the result page shows where the routes agree or differ.
- Are face shape categories the same on every site?
- Not exactly. Most guides use the same broad set, but the boundaries move: some split out an inverted triangle, some treat oblong and rectangle as one category, and some use a different name for the same outline. That is why this site publishes its own rule set and the reference each label is centred on, rather than asking you to trust a diagram.
Photo analysis and privacy
What leaves your device, what the detector measures, and why a photo can be refused.
- What happens to my photo after I close the tab?
- Everything is gone. Your photo existed only in your browser's memory, which the operating system reclaims when the tab closes. There is no server-side copy to delete, because there is no server-side copy.
- Is the analysis really running on my device?
- Yes. The landmark model is compiled to WebAssembly and runs inside your browser. The image is decoded locally, the landmarks are detected locally, and no network request carries the photo. You can watch the Network tab while a photo is analysed if you want to check that claim yourself.
- What does the photo analysis measure?
- It locates 478 facial landmarks, then reduces them to five readings: face length relative to cheekbone width, forehead width relative to cheekbones, jaw width relative to cheekbones, the angle at the jaw corner, and chin taper. The classifier compares those readings with seven reference shapes and returns the closest one, plus the margin.
- Why does head tilt or camera distance change the result?
- A face is measured in the image, not in three dimensions. Tilt changes the apparent angle at the jaw, a close camera stretches whatever is nearest the lens, and a low or high camera changes which part of the face that is. The detector asks for a level, straight-on photo because those distortions move the same landmarks the ratios are built from.
- What makes the detector refuse a photo?
- It refuses when the face is too small in the frame, turned too far to one side, tilted, not level with the camera, or missing key points. It also refuses a photo with no face or with several faces. The fix is usually a closer, straight-on shot with even light and hair pushed back.
- Do glasses, a beard or a fringe affect the reading?
- They can. Frames can cover the outer eye corners, a beard can blur the jaw line, and a fringe hides the hairline the face length is measured from. The analysis may still run, but expect a wider margin or a lower confidence. If the result matters to you, take a second photo and keep the frame or hair away from the measured points where you can.
- Do you train models on my photos?
- No. The site never receives your image, so there is no photo to train on. The landmark model is a pre-trained browser build; your photo stays in memory for the length of the analysis and is not added to any dataset.
Quiz and tape measure
Two routes that work without a photo, and how they cross-check the first.
- How is the quiz different from the photo analysis?
- The quiz asks seven questions about what you see in a mirror; it does not use a camera or a photo. The photo analysis measures 478 landmarks. They are different inputs rather than a better and a worse version of the same test, and the result page shows where they disagree.
- Which four measurements does the tape calculator ask for?
- Face length, forehead width, cheekbone width and jaw width. You need a soft tape measure, or a piece of string and a ruler, plus a mirror. The calculator derives the jaw angle from those widths and treats chin taper as unavailable, which is why the result page caps its confidence and says the cap was applied.
- Which method should I try first?
- Use the photo analysis if you have a straight-on, well-lit photo and want the fastest read. Use the quiz if you have no camera or want a route that needs no photo at all. Use the tape calculator when you want an independent second measurement, especially if two shapes are close.
- What if two methods give different shapes?
- That is useful information, not a failure. The result page shows which features the two paths disagree on, and borderline faces genuinely sit between two reference shapes. Retake the photo if it was tilted or small; otherwise compare the margins and treat the label as a close call rather than a verdict.
The attractiveness test
What the 0–100 index measures, and the questions it deliberately does not answer.
- What is the AI attractiveness test actually scoring?
- Seven rows: left-right balance, facial thirds, facial fifths, a declared golden-ratio reference, eye-corner balance, jaw balance, and eye-to-nose spacing. Each reading is compared with a declared reference, and the 0–100 index summarises how closely the set matches. It is not a panel of human raters, and it does not predict how other people will see you.
- Does a higher score mean a better-looking face?
- No. The index measures closeness to declared geometric references, not attractiveness as people experience it. A higher number means the measured set sits closer to those references; it does not mean the face is better, and a lower number does not mean it is worse. The five bands describe that distance, not a grade.
- Why is there no percentile or ranking?
- A percentile needs a representative distribution of the population you are comparing against, and this site does not have one. The calibration set is a working reference set, not a population sample, so a top-percent label would add precision the data cannot support. The page reports the band and the row-level differences instead.
- Is the score stable across photos?
- The method is reproducible for the same photo and the same method version: the same image should give the same result. A different photo can give a different result because head angle, camera distance, light and expression move the landmarks. Treat the number as a reading of one image, not a fixed property of your face.
- Is the reference set a representative sample?
- No. The references are declared values and published proportions, not a population average, and the calibration set is not a representative sample. That is why the page does not compare you with other people or label a score as typical. It compares your readings with the references it names.
- What leaves my browser when I run the test?
- Your photo, landmarks and row readings stay in the browser. They are not sent to a server, and there is no stored copy to delete. The site does load two page-view counters on every page; they receive the page address and basic device information, not your photo or result. The privacy page lists both counters and what each receives.
Styling and recommendations
Hairstyles, glasses and accessories — and who pays for what.
- Is one face shape better than another?
- No. Face shape isn't a score. There's no better or worse shape — only styles that work with the one you have. Every page on this site is written on that basis.
- Do you sell glasses or hair products?
- No. We don't sell glasses or hair products, and the recommendations on this site aren't influenced by any brand. There are no affiliate links and no paid placements on this site today; if a retailer link is ever added, it will be disclosed next to the link and on the /about page.
- Are any product recommendations paid?
- No. There are no sponsored posts, paid placements or affiliate links on the site today. The styling pages describe how lines and proportions interact; they do not link to a shop. If that changes, the disclosure will sit next to the link and on the about page, and the link will be marked as sponsored in the page source.
- Which hairstyles suit my face shape?
- Each of the seven shape guides has a hairstyle section that explains which cuts add length, width or angles and why. The lists are options, not rules: a cut that is not listed is not a mistake. Start from your shape's guide and use the reasoning to judge a style you already like.
- How do I choose glasses that fit?
- Two separate questions: the frame has to clear your cheekbones, and the frame shape is a styling choice. The frame-size calculator takes the lens width and bridge printed inside a pair, plus your pupillary distance, and checks the fit. The shape guides explain which frame lines contrast with your outline.
- Do hats, earrings and necklaces depend on face shape?
- Not in the same way. Hat size comes from head circumference, necklace length from a drop measured on your neck, and earring size from your jaw line at your cheekbone width. Each tool asks for the measurement it actually uses, rather than sending you through a face-shape quiz first.
Site, data and corrections
Cost, analytics, open data, and how to challenge a claim.
- Why don't you advertise an accuracy percentage?
- Because the accuracy figures attached to face-shape tools are self-reported — no stated sample, no stated method, nothing to check them against. We publish our thresholds and your measured ratios instead, and we will publish a real confusion matrix once we have a labelled validation set large enough to mean something.
- Is the whole site free?
- Yes. Every tool on the site is free to use, with no account, no sign-up and no paywall. There is nothing to unlock and no premium tier; the analysis runs in your browser either way.
- What data do you collect?
- Nothing about your face. Photos and landmarks stay in your browser; the site never receives an image. Two page-view counters, Cloudflare Web Analytics and Google Analytics, receive the page address, arrival time, referrer and basic browser information. Google Analytics sets a first-party cookie; the privacy page explains both counters and how to block them.
- Can I reuse the rule set and the data?
- Yes. The face-shape rule set and the dated tables the articles publish are in a public GitHub repository under CC BY 4.0. You can reuse them with attribution; the open-data page lists the repository, what is included and what the data does not prove.
- Who writes and reviews the site?
- The long-form pages are written by the Know Your Face Shape team. They are not yet reviewed by a licensed cosmetologist, and the byline says so rather than inventing a reviewer. If you are a licensed cosmetologist or a craniofacial researcher and spot something wrong, the contact page reaches us.
- How do I report a mistake?
- Send the page address and the sentence to contact@knowyourfaceshape.com, with a source if you have one. The about page lists the references the method is built on, which is the fastest way to show where a claim and its source disagree.
Still not sure?
Send the page and the sentence that looks wrong to contact@knowyourfaceshape.com. If the question is about your own result, include the shape and the method that produced it.
Contact