Parental control menus move constantly, so any guide site is either maintained or quietly wrong. This is the record of what we have changed and when, including the things we got wrong. It is published because a verification date you cannot check is just a claim.
40 entries13 corrections
Correction
We had the Irish Internet Hotline’s name wrong, and undersold what they do
Corrected by the organisation itself after we sent them the page to check. Two things. Their operating name is the Irish Internet Hotline, not Hotline.ie, which is their reporting address rather than what they are called. And we had described them as somewhere you send a report, which undersells it: they assess every report and work with internet service providers and other organisations to get illegal content taken down. For a parent deciding whether a report is worth making, that is the difference between filing something and getting something removed, so it belonged on the page. Both are fixed on the Irish crisis page.
Correction
The Roblox guide said no setting touched private servers. There is one.
The note on the chat step told parents that being shut out of a private server is the version of Roblox bullying they hear about most, and that no setting here touches it. The first half is right. The second half was too broad. Roblox does have a private server control, at Settings, Parental Controls, Settings You Manage, Visibility and private servers, and we had not covered it at all. It has its own step now. The chat note has been narrowed to say what is actually true, which is that none of the chat settings reach it, and the new step is honest that the server control governs who can get into your child's server and does nothing about your child being kept out of someone else's. That is the part that hurts, and no setting anywhere fixes it.
Change
Roblox: trading, Robux transfers and four more controls we had missed
Comparing our guide against Roblox's own parental controls page turned up controls we were not covering. The largest gap was money moving sideways. Trading and Robux gifts are how a child gets talked out of something they paid for, and a spending limit does nothing about either, because it stops them buying Robux rather than giving away Robux they already have. That is a step now, with inventory visibility and Roblox's own transfer caps of 500 Robux a day and 1,000 a month. Blocking specific games, the request-and-approve route for a blocked game, and the sensitive issues toggle have been added to the content maturity step. The friends step now carries the official menu path and the fact that a trusted friend request expires after 24 hours, which to a child looks exactly like a refusal.
Change
Some steps now link to the platform’s own video
A written menu path is the fastest route for most people and useless for some. Where a platform has published a short video of someone tapping through the exact setting a step describes, that step now links to it. Five Roblox steps have one so far. The rule, which the build enforces rather than trusting anyone to remember, is that a video is only ever the platform's own, and its label has to name the platform. A third party's explainer goes stale the moment they stop posting, and it puts a stranger's face on a page a parent is trusting for their child's safety.
Change
SafeStart installs now, and works with no signal
The site is now an installable app. Add it to a home screen on Android or a phone and it opens full screen with its own icon, and every guide keeps working with no signal at all. That last part is the reason to bother: a parent setting up a child’s tablet is often doing it in a back bedroom on one bar, or on the very device they are about to restrict. All twenty-nine guides are readable offline, and so are the crisis pages, which are the ones most worth having when everything else has gone wrong. One rule was written into the code rather than left to good intentions. The crisis pages are never served from a cache while any network exists. Everything else on this site can be a few days out of date and a parent still ends up better off, but a stale crisis page sends a frightened person to a reporting route that has moved or an organisation that has been renamed, and that is the one failure this site cannot have. So those pages are fetched fresh every single time, and the stored copy exists only for the case where there is no connection at all, because a slightly old crisis page still beats a browser error at two in the morning. The build fails if anyone ever changes that, and the test proves it by editing what the server returns between two page loads and checking the page follows. The icons were redrawn too. They were still the old navy TrustRaise shield, left over from before SafeStart had its own colours, and they now use the same drawn shield as the header, on the same forest green. They carry new filenames, because the old ones are served with a one-year cache and overwriting a name would have left people looking at the old icon well into next year.
Change
What the settings do, and do not do, about bullying
Someone asked what the site says about cyberbullying and the honest answer was: less than it should. The crisis page has had a proper route for a while, with advice to use the platform’s own bullying category rather than a general report, and the point that bullying between classmates is a school matter with an online surface. But the guides themselves said almost nothing. One line in Snapchat’s risks, about messages disappearing taking the evidence with them, and that was it across twenty-nine guides. So fourteen steps now carry a note, and only the fourteen that genuinely decide who can reach a child: chat settings on Roblox, comments on TikTok and Instagram, direct messages on Discord and Twitch, group adds and blocking on WhatsApp, contact settings on Snapchat, the console communication settings, Android contacts and iMessage Communication Limits. Every one of them names something the setting does not reach, and that rule is enforced by the build rather than left to good intentions, because a note saying only that a setting helps is worse than no note at all. A parent who reads that and stops reading has been misled. None of these settings stop a classmate being cruel. What they change is who is able to reach a child and what gets through, and several of the notes exist specifically to say so: closing Discord DMs does nothing about a server full of classmates, Android’s contact controls miss every app where it actually happens, and Communication Limits only work on people who are not already in the contacts. The build also fails if these ever spread past a quarter of all steps, because at that point nobody reads them. The crisis page gained one thing too, and it is the part most people get wrong: save it before you block. Blocking can put the whole conversation out of reach, Snapchat messages delete themselves, and WhatsApp blocks screenshots of View Once entirely. Photographing the screen with a second phone is the only method no app can block or notify about, and it works on a console as well.
Change
New typeface, redrawn icons, and the first people on the site
Readers kept describing SafeStart as corporate, and pushed on it the answer came back as two things rather than a vague feeling. The first was the typeface. Inter is the face of roughly every piece of software a parent meets at work, so it was arriving before the words did. It has been replaced by Atkinson Hyperlegible Next, drawn by the Braille Institute so that no letter can be mistaken for another, which is why the lowercase l has a tail and the capital I has serifs. It is warmer and a little odder than Inter without being soft, and on a site read in a hurry by worried people on phones, picking a typeface for legibility is a decision that stands up on its own. One variable file covers every weight, which is 53KB against the 97KB it replaced, so the site also got slightly faster. The second was that everything on the page was interface. Every icon has been redrawn: no rectangles, no perfect circles, every straight run is now a shallow curve and no two corners share a radius, on a slightly heavier stroke. The settings paths were set in monospace, which made a list of taps look like a line of code, and a parent who does not write software reads a code block as not being for them. Same words, ordinary type, arrows doing the work. And three drawings have been added, which are the only place on the site a person appears: two figures on the home page, a phone on the About page, and one that shows up on a finished plan and nowhere else. They have no faces, because a drawn face is a specific child and the parent reading this has their own. Not a single word of guidance changed anywhere on the site, which was checked page by page rather than assumed.
Change
Every guide now says what happens when the menu names do not match
A reader asked which version of Android or iOS the steps were written for, and the honest answer was that nothing on the site said. Every guide carries one line about it now, sitting with the checked-on date because that is where a parent is already asking whether this is current. It deliberately names no version number. A number reads as precision but it is a promise with an expiry date, and the moment it is stale a parent on a newer phone decides the guide is abandoned while a parent on an older one decides it was never meant for them. Both would be wrong, because the setting almost always survives the rename. So the line says menus move between versions and between phone makers, and that a name which does not match is a reason to look nearby rather than to assume the setting is gone. It is in the markup as well as the app, so a reader with no JavaScript gets it too. The build now fails if a version number or a year ever gets added to that line, which is the sort of decision that quietly erodes otherwise.
Change
A Samsung Galaxy guide, because a Galaxy has two app stores and two browsers
A Galaxy is an Android phone and Family Link supervises it exactly as it supervises a Pixel, so the Android guide was never wrong. It was incomplete. A Galaxy ships with the Galaxy Store alongside Google Play and Samsung Internet alongside Chrome, and Family Link reaches neither of them. A parent who set up download approval in Family Link and filtered Chrome has a child who can install from the other store and browse in the other browser. Samsung has its own controls that do cover both, through a Samsung child account, and they are what this guide sets up: Galaxy Store age ratings and download approval, and Samsung Internet web content. There is a genuine win in there. Samsung states that secret mode is not available while using a Samsung child account, so the private browsing everyone assumes cannot be switched off is switched off by the account itself. The guide also covers Samsung Kids for younger children, including the documented route back in when you have forgotten the PIN, and it ends with Secure Folder, which is the one thing here with no parent-side control at all. That step says so rather than pretending. Picking Samsung Galaxy in the device list now gives you the Android steps and then the Samsung ones, in that order.
Correction
Clicking SafeStart from the crisis page left you looking at nothing
Reported by a reader, and the worst version of it had been live since the crisis pages were built. Tapping the SafeStart mark in the top left from the Get help page, or from About, the changelog or the feedback page, produced a page with a header, a footer and a hole where everything else should be. The only way out was reloading the whole site. The cause: the crisis pages deliberately do not run the app at all, because they have to work when every other part of this is having a bad day, which means the guide data is never loaded on them. But the SafeStart mark was a button rather than a link, and its click handler called the code that builds the home page regardless. That code cleared the page first and then failed on data that was not there. It is a plain link to the home page now, which works with or without JavaScript and never depends on anything having loaded. On pages where the app is running it still does the fast in-page version, and modified clicks are left alone, so open in a new tab and middle click work for the first time. There is also a guard so that any other route into the home page with no data loaded sends you to a working page instead of blanking the one you are on. Six assertions cover getting home from the crisis page, About, the changelog, feedback, a guide and a plan.
Change
There is a way to tell us we are wrong now, and one heading has stopped being clever
The About page had a section called 'Why you should not just trust us', which was a line that sounded good and told you nothing. A reader said as much. It is 'How to check we are right' now, and it says what it means: every guide shows when it was last checked and links to the platform's own page, so you can read the source yourself rather than taking a date on faith. The more useful change is that 'Found something wrong?' now goes somewhere. There is a feedback form at /feedback/, and a button on every guide and plan that opens it carrying the page you were on, so you do not have to describe where you were. A message is required, an email address is optional and only used to reply to you, and the page says plainly that your message is emailed to us rather than stored here. Please leave your child's name out of it. Two things about how it is built, since a form that quietly fails is worse than no form. If sending is not configured, the page says so and hands you a prefilled draft in your own mail app instead of pretending to work. And there is a hidden field on the form that people never see and automated scripts fill in every time; anything that arrives with it filled is dropped, and answered as though it succeeded so the sender learns nothing. The whole route is tested, including that a newline cannot be smuggled into a mail header. One thing changed straight after it was built, and it was worth catching: the feedback button first shipped with a warning triangle and the words Something wrong?, which put it in exactly the same visual language as the Get help now button in the header. Those are not the same kind of thing. One is for a child who is in danger right now, the other is for a menu that has been renamed since we wrote about it. Sharing an icon between them teaches people to discount the triangle, and the triangle is the one piece of iconography on this site that has to keep meaning what it says. It is a pencil and the word Feedback now, in the informational colour rather than the warning one, and a test fails if the two ever converge again. Reader corrections are the single most valuable thing this site gets. Three of the last four changelog entries started with someone writing in.
Change
A reader asked what this actually is, and they were right that we never said
Seven pieces of feedback from someone opening the site cold. Three of them turned out to be the same thing. One reader clicked a link and could not tell what SafeStart was or why they would use it. Another asked whether their child would need a browser on their phone to create or activate their account, because their 11 and 13 year olds are not allowed browsers. A third finished a plan and asked what happens next, and whether parents get alerted if something is amiss on their child's phone. Three different questions with one cause: the site never said what it was. The home page opened with 'Let's set this up together' and went straight into a picker, which only makes sense if you already knew. It now says in its first sentence that SafeStart is a free guide to the parental controls already built into your child's devices and apps, and there is a new About page for the rest. That page says the things we would rather be asked than assumed. It is not monitoring software, nothing keeps running after you finish, and no alert is ever coming from us. It does not read anyone's messages, and neither does Apple's or Google's parental control whatever you may have been told. There is nothing to install, no account, and your child never visits the site. It also states what is counted, that the plan page is deliberately left out of that because its address describes a specific child, and that a question typed into Ask SafeStart goes to an API to be answered. The plan now ends with a section on what to do next rather than stopping dead after the last step, including the most useful thing you can do: try to install an app on your child's device and see whether the request actually reaches you. Building that page turned up a bug that had been there far longer. The app only knew that the crisis pages were static, so every other static page fell through to code that clears the page and paints the picker over it. The changelog had been doing that since it was built. A search engine saw the changelog; a reader with JavaScript on saw the front page. Both are fixed and tested. Three smaller things from the same reader. Build my plan used to leave you at the bottom of your own plan because the page never scrolled back up, and fixing it turned up that the site's smooth scrolling would have made the jump an eight hundred millisecond glide, so it is instant now. The site said 'they' throughout without ever saying who 'they' was; the picker asks about your child by name now. And the question above the three device buttons has been rewritten again, because 'in your plan' means nothing before you have seen a plan. One thing added rather than fixed. The reader wanted, for Android, what the iMessage guide gives on iPhone. There is a real control there and we did not cover it: Family Link lets you decide who can call and text your child, with approval for anyone new. It is in the Android guide now, with the limit Google states plainly, that it only reaches Google Messages, Samsung Messages and Google Dialer, so it locks the front door and leaves WhatsApp and the rest exactly as they were. Worth saying clearly, since the question was about monitoring: no parental control on either platform lets you read your child's messages.
Correction
A reader worked through the Fire tablet guide with the real dashboard open, and found four things
The most useful kind of feedback: someone followed the guide with the actual settings in front of them and wrote down every place we did not match. The worst of it was a bug rather than a mistake in the writing. If you opened Ask SafeStart from a plan and then closed it, the button never came back, for the rest of your session, with no way to get it. It worked fine on a full guide page, which is why we never caught it: the code that put the button back only knew about guides and nobody had tested it on a plan. It is fixed, one function decides whether that button belongs on screen now instead of six different places setting it by hand, and there are tests for both surfaces. The same reader asked for a way to park the chat rather than close it, because everything behind it is blurred and they wanted to read the step the answer was about. There is now a minimise button next to the close, and once you have asked something the button reads Back to your question rather than Ask SafeStart. Two Fire tablet corrections, both from screenshots of the live dashboard, which is better evidence than a help page. Amazon renamed the age filter: it is Content Maturity Filter & Languages now, not Adjust Age Filter, and it is a range with a lower and an upper age rather than a single age, so the instruction to set it to their real age was not quite right either. And our path for the browser setting did not exist. We said Parent Dashboard, Settings, Fire Tablet Web Browser. The reader looked on desktop and in the app and could not find it. Amazon's own wording is that browsing controls are customised for each child in the Parent Dashboard, under that child's settings, so that is what the guide says now. Two wording fixes as well. Why this one? has become Why this matters, because the old phrasing sounded like it was comparing options rather than giving a reason. And the question above the three device buttons was not asking anything: it said THE FIRE TABLET ITSELF and left you to infer the question, with one reader wondering if it meant a device fresh out of the box. It is a written question now, with a line underneath explaining what device settings are and why most people want them.
Change
New colours, because readers kept saying it did not feel welcoming
Plenty of people said the site was useful and several said it felt cold. Looking at it properly, the cause was not the blue on its own. Every colour on the page was cool: the background had a blue cast, the headings were navy, the accent was an electric blue used for banks and dashboards, and there was no warm tone anywhere. A page with no warmth in it feels like an office no matter how good the content is. Even the shadows were tinted navy, which is a small thing that does more than you would think. So SafeStart now has its own palette rather than TrustRaise's. The ground is a warm blush, and the page speaks with three colours instead of one. Forest green means you do something here: buttons, links, step numbers. Clay means a warning, or a setting your child can undo. Bluebell means explanation rather than instruction, so the reasoning and the plain-English translations sit in it. Two soft fills, butter and blush, carry the recommendations and the cautions. The point is not that there are more colours, it is that three different kinds of information now look different. Before this, a recommendation, a warning and an explanation were all the same blue box. Dark mode was rebuilt rather than tinted. Its ground is a deep warm brown rather than the old navy, because dropping warm accents onto a cold dark would have made the toggle feel like two separate products. Every colour lifts to match: forest becomes sage, the button flips to dark text on a light fill, and the three voices keep their jobs. Two things worth saying about how this is kept honest. The colour tokens are now named after what they mean rather than what they look like, so nothing in the stylesheet is called blue while holding a green, and the next person to touch this can see the intent. And the build now checks 24 colour pairs across both themes against WCAG AA and fails if any of them slips. It caught two problems while this was being written, one of them a colour that was one hundredth of a point under the line. Warm palettes are easy to drift into unreadable, and a warning nobody can read is worse than no warning. The TrustRaise connection is still there in the shield, the typeface and the byline. It is just no longer carried by the colour.
Correction
Nine things the weekly check caught, and one of them we had flatly wrong
The Monday job re-read all 27 guides and flagged nine things. Each was checked by hand against the platform's own page before anything was changed here, and one was rejected for lack of evidence. The one that matters most is iMessage. We said, without qualification, that Communication Safety runs on the device and that Apple never sees the photo and never learns that anything was found. The first half is right and Apple says so. The second half is not. If your child chooses to report the content in iMessage or FaceTime, Apple's own page says the sensitive content is sent to Apple for review, and Apple may pass reports to law enforcement and act on the sender's account. That is useful for a parent to know rather than a caveat, so it now sits in the guide as its own point. Apple also renamed the strictest web setting. What used to be Allowed Websites Only is now Only Approved Websites, on iPhone, iPad and Mac. The job spotted it on two of the three; the iPhone guide had the same stale name and came back unclear that week, so we fixed all three. Nintendo changed two things at once on Switch 2, and we had one of them backwards. Posting to social media was removed from the console, so that restriction cannot be set there any more. And GameChat, which is built into Switch 2, is inside parental controls rather than outside them: Nintendo requires a parent to be involved in setting it up and using it for anyone under 16 or on a supervised account, and permission is needed every time a younger player wants video chat. Our guide said flatly that voice chat sits outside these controls. That is still true of the separate phone-app voice chat for 13s and over, so the guide now separates the two. Smaller ones. Instagram's Sleep Mode runs 10pm to 7am, not 8am as we said. Microsoft moved its family page, and their own steps now say to sign in at account.microsoft.com/family. Discord turned out to be stronger than we implied: a teen cannot switch the sensitive content filter off at all, and the Show option is not offered to them, so the only real choice is whether it blurs or blocks. The YouTube naming flags were already fixed last week. One finding was rejected. The job reported that changing Discord's message request settings now needs an adult age check. Discord's own page on messaging does not say that, and does not describe teen defaults at all, so we left our wording alone rather than assert something we could not confirm. The PlayStation hand-check is still outstanding, because PlayStation blocks automated readers and that page has to be opened by a person.
Correction
Every setting now says whether your child can just switch it back
This one started with YouTube telling us we had autoplay wrong. We had sent parents to the autoplay switch inside the child's own YouTube settings, which the child can undo, when the one that holds is on the parent's account. Then building the Spotify guide turned up exactly the same thing: the explicit filter everyone points at lives in your child's settings, and there is no parent-side lock for it. Two is a pattern worth checking, so we checked. Across the site, 53 steps send a parent into an app's own settings screen. We sampled fourteen of the ones most likely to be affected, and all fourteen were silent on whether the child could reverse them. Discord's privacy settings. WhatsApp's four privacy controls. Snapchat's Ghost Mode and contact settings. Twitch chat filters. Instagram's messages, privacy and time settings. Roblox's age check. TikTok's privacy and contact settings. Every one of them left the most important question unanswered, which is the failure that actually matters here: a parent works down the list, ticks the checklist, and believes they are finished, when some of what they set will be gone by the weekend. So every settings step on the site now carries one line saying who holds it, in the same words on every app, the way the chat settings already do. There are five answers. You set this and your child cannot change it back. Your child cannot loosen this without proving to the platform that they are an adult. This lives in your child's own settings, so they can change it back. Either of you can change it and neither is told when the other does. And it is held by the device passcode rather than the app. The count across 136 settings steps: 110 you hold, 20 your child holds, 4 shared, 1 the platform locks, 1 the device. Two of them change hands at a birthday, and those are the ones worth knowing about most. Roblox hands chat and voice to your child at 13. Instagram lets a 16 year old loosen what a 15 year old needs your permission for. Those now give a different answer depending on the age you have picked. The reassuring answers matter as much as the warnings. Most of this site describes controls that genuinely hold, and until now we never said so. One thing verified on the way: Discord teens cannot turn the sensitive content filter off at all, only choose whether it blurs or blocks. The build now fails if anyone adds a settings step without answering the question.
Change
Spotify added, and the explicit filter does less than everyone thinks
A parent asked how to stop explicit lyrics and videos. The short answer is that the switch every article points you at covers songs and stops there. Spotify's own guidance for parents says the filter applies to songs tagged explicit by creators and rightsholders, and not to podcasts, audiobooks, or videos tagged 18+ by rights holders. Podcasts are where most of the strong language on Spotify actually is, and that switch does nothing to them. Two more things worth saying plainly. On an ordinary account the filter lives in your child's own settings, so it is theirs to switch back, and there is no parent-side lock for it. And video is a separate control again, covering music videos and Canvas, the short looping visual behind a track, which is the other half of what the question was about. For an under-13 there is a much better answer than any of this, and it is fairly new. A managed account is music only, so podcasts and audiobooks are absent rather than filtered. Explicit content is blocked automatically on every new one, video and Canvas are off by default, the profile cannot be searched, there are no purchases, and it sits behind a parental controls PIN, so unlike the ordinary filter your child cannot undo it. It runs on the free tier with ads as well as on Premium Family. It reached the United States and the UK on 15 July 2026 and Canada and Ireland on 28 July, so all four countries here have it. One good surprise: a young listener stays on a managed account until 18, not 13, and keeps playlists, saved music and recommendations when they graduate. The uncomfortable part is the teenagers. There is no managed account for a 13 to 17 year old, and apart from video on a Family plan there is no Spotify setting they cannot reverse. The guide says so rather than pretending otherwise, and treats those years as a conversation instead of a control panel. One thing we could not confirm: whether converting an existing account to a managed one can be undone. Spotify says the young listener will immediately lose access to some content and features, and is silent on going back, so the guide calls it a decision rather than an experiment.
Change
A third option for the parents who already set it up, badly
Several parents said the same thing: they had already set devices up, got it wrong, and every route on this site assumed a clean start. The device chooser now has a third button, I set it up wrong, next to Set it up too and Already done. It does not send you somewhere new. It gives you the same steps, with a note on each one a previous attempt can block, saying how to undo that first. Twenty-two steps carry one. What it leads with matters more than the notes though. The instinct when you have made a mess is to start over, and starting over is usually the wrong answer and occasionally an unrecoverable one, so fix mode opens with what cannot be taken back before it tells you to do anything. On PlayStation a date of birth cannot be changed at all, you cannot revert a child account once it has been upgraded, nothing transfers to a new account, and child members cannot be removed from a family once created. A Nintendo Account date of birth cannot be changed either. TikTok has no self-serve correction, only an appeal that opens after they restrict the account, and an under-13 account is deleted on day 120, or day 30 in the United States. On a Fire tablet the lock screen PIN resets with your Amazon password while the parental controls password means a factory reset, and most people do not know which one they have forgotten. On Google, correcting a birthday downward starts a 14-day clock and then deletion at 30 days. The good news is the larger half. Two things everyone repeats are out of date. An adult Apple Account can now be converted to a child account inside Family Sharing, keeping purchases, photos and Messages. Google supervision can be added to an account your child already has, and since the change a 13-year-old can no longer switch it off alone, because parent approval is needed until 18. Microsoft converts with the gamertag, Game Pass and Minecraft entitlements intact. Instagram, Roblox and Netflix all correct in place. Four things we could not confirm against an official page are still in the guides, but they say plainly that we could not confirm them: the Xbox passkey reset, whose page is script-loaded and unreadable, resetting a Screen Time passcode on a Mac, how many times Instagram lets you change a birthday, and whether an existing WhatsApp account can be converted to a parent-managed one. A test fails if anyone ever edits the hedge out of those four.
Correction
We had the YouTube guide wrong in several places, and someone at YouTube told us so
The specific complaint was autoplay. We sent parents to the autoplay switch inside the child's own YouTube settings, which the child can switch straight back. The one that holds is Disable autoplay in Family Center, on the parent's account, and YouTube's wording is explicit: your child can't turn on Autoplay when this setting is on. It also ships off on kid accounts already. Checking that turned up more. We said an under-13 can only use YouTube Kids, which has not been true for years. A supervised kid account is, in Google's words, a parent-managed version of regular YouTube for children under 13. In fairness to us one of Google's own live pages still says the old thing, and we followed it. We had missed supervised teen accounts entirely, which are a different feature with no content filter, needing the teen's agreement, and endable by either side. We said supervised accounts can read comments but not post them, which is right for the two higher content settings and wrong for the lowest, where they cannot read comments at all. The block-channels steps described the YouTube Kids route rather than the supervised one, which is a channel page, About, Report user, Block channel for kids, done from your own account. We said supervision ends automatically at 13, and nothing happens on the birthday by itself. And we said YouTube has no real time limit, which undersold the YouTube Kids timer that locks the app. The biggest gap was structural: Family Center is where nearly all of this lives and we never named it once. It now has its own step. Two things we are not going to pretend are tidy. Google uses two sets of names for the same three content settings, Older kids and Explore, Teens and Explore more, Older teens and Most of YouTube, and different Google pages use different ones, so the guide gives both. And adding supervision to an account a child already has deletes their subscriptions, playlists, comments and purchased videos, which Google states plainly and which no parent expects, so it is now the last thing you read before you start.
Correction
Our Roblox chat step never mentioned voice, and used the old names for everything else
A reader put a plan side by side with Roblox's own parental controls page and found voice chat missing from our guide entirely. Checking that turned up a second problem: Roblox's Safety Center and its Help Center use different names for the same settings, and we had followed the Help Center. So Experience chat was Game chat here, and Chat and party with friends was three separate items. Both are fixed, with the old names in brackets so readers of either page can follow along. Voice now has its own step, because there are two voice features with two separate off switches and turning one off does nothing to the other. Also added: the age check, which since 2026 is the gate on all chat at any age, and Gameplay Coordination, which almost no guide covers. Two things we now say plainly that we did not before. None of these settings reach chat that a game's creator built themselves, which Roblox states twice on its own page. And at 13 your child takes over their own chat and voice settings.
Change
Every recommendation now says what it means, in the same words every time
From a reader: set Roblox chat to 'All chat off', then meet PlayStation's 'Restricted, no messaging or chat', and you have to work out for yourself that these are the same thing. Every platform names its settings differently and the translation is left to the parent. So the platform's exact wording stays, because that is what is on your screen, and underneath it sits one sentence that is word for word identical across every app. 'Only people already on their friends list can reach them. Strangers cannot.' means the same on Roblox as it does on Xbox, Snapchat or an iPhone. Twelve settings carry it so far, and the build fails if two apps ever drift to different wording for the same thing.
Change
Apple Watch guide added, including the thing Apple buries
Asked for by a reader. The feature is called Apple Watch For Your Kids now, renamed from Family Setup, which is why older guidance uses a different name. The guide leads with checking it will work at all, because Apple's own pages disagree about which models qualify and the carrier list is short. In the UK that is EE. Ireland is not on Apple's list at all. The buried detail worth knowing before you buy: Fall Detection does not work for anyone under 18, and Apple says so only in a list at the bottom of the setup page. A lot of people buy these for a child partly on the strength of that feature. Crash Detection does work. Sleep tracking, cycle tracking, blood oxygen and international roaming do not.
Change
The em dashes are gone
Ninety-one of them, mostly in the source links. Small thing, but house style is house style.
Change
We turned on visitor counting, and left it off the one page that matters
SafeStart now counts page views using Vercel's analytics. It is worth saying exactly what that means, since a site about other people's privacy should be straight about its own. No cookies are set. Visitors are identified by a hash of the request rather than anything stored on your device, and that is discarded after 24 hours. What gets recorded is the page, the referrer, a rough location, and the browser and device type. The script is served from this domain, so the site still makes no third-party requests at all. One deliberate exception: the plan page is excluded. Its address contains the age you picked, the device, and the list of apps, which together describe a specific child. That is aggregated and anonymous either way, and it still is not a good enough reason to send it anywhere. The build refuses to put the script on that page, and a test fails if anyone changes that.
Correction
Ask to Buy does not stop a child re-downloading an app, and nor does Google's version
A reader told us her sister got apps from the App Store for a long time before their parents noticed. There is a documented reason. Apple's own wording: if a child redownloads their own purchase, installs an update, or uses a redemption code, you will not receive a request. So an app approved once, or installed before Ask to Buy was switched on, can be deleted and pulled back from their purchase history any time, silently. Google has the same hole and states it just as plainly: previous downloads and content shared through Family Library will not require your approval. Neither company mentions this anywhere near the setting itself. The iPhone, iPad and Android guides now lead with it, and point at the separate control that does cover redownloads.
Change
A guide for the accounts, because everything here used to start too late
Two readers asked the same question a week apart: children end up with their own iCloud account, so does that mean a child needs an email address? Every guide on this site started after the account already existed, which is too late. There is now an Accounts and email guide, and it leads every plan. The answer to the question, from Apple's own page: no. Creating a Child Account lets you pick their own email address, a suggested @icloud.com one, or just a Game Center nickname, and the nickname route means no email at all. Google is different: a supervised account normally needs a Gmail address, with one narrow exception for an under-9 on a Pixel Tablet. The guide also covers what the account switches on without asking, and why a child on a parent's account receives the parent's messages and call history, which Apple documents itself.
Change
iMessage and FaceTime have a guide, and every device says what is already on it
iMessage is on every iPhone and iPad, has no minimum age of its own, and for a lot of children is the first place a stranger can reach them. Nobody installs it, so nobody thinks to set it up. It now has its own guide, covering Communication Limits, which genuinely block rather than warn, and Communication Safety, which parents widely assume will alert them and does not. It detects nudity on the device, blurs it, and for under-13s prompts the child to talk to you. It never tells you anything. The same blind spot applies elsewhere, so every device guide now lists what ships on it. On Android that includes YouTube, and it is the full version rather than YouTube Kids, which is a separate download.
Change
HBO Max added, with the thing that catches people out
Asked for by a reader who uses it for younger children. The catch is worth stating up front: a Kids profile is not a lock. Kid-Proof Exit is off by default, adult profiles have no PIN until you add one, and on phones, tablets and computers the app reopens on the last profile used with no Who's Watching screen. So out of the box the profile switcher is one tap from the full library, and it takes two separate opt-ins to close that. Two content notes, because they change whether it is worth subscribing: Sesame Street moved to Netflix in 2025 with worldwide premiere rights, and Studio Ghibli is on HBO Max in the United States, where GKIDS holds exclusive US streaming rights. HBO Max is not available in Canada. It launched in the UK and Ireland in March 2026, where Sky and NOW customers often already have the ad-supported tier included.
Correction
Quick setup was hiding the reason for every single step
A reader said she did not understand why we suggested pausing YouTube history, since history is how you see what your child has been watching. She found the answer by asking the chat, which told her it stops the recommendations chasing whatever they clicked once. Looking into it, the reason was never on her screen at all: Quick setup dropped the why from every step, and Quick setup is the default. A parent told to tap something with no reason given is the one most likely to switch it back a week later. The reason is now one closed line on every step, so quick still means quick, and it opens with a tap. Learn as we go is unchanged.
Change
Twenty-four steps now say what the setting costs you
Having a reason on the page is only useful if the reason is honest about the trade. So we went back through every step where the setting either looks wrong or takes something away, and said so plainly: Ghost Mode hides your child from you as well as from strangers, turning Roblox chat off can mean playing alone, going private on TikTok costs the views that are the point of the app for a lot of teenagers, Ask to Buy pings you for free apps too, automatic web filtering will block the odd homework page, and a Minecraft Realm is a paid subscription, which is the real reason families end up on public servers instead.
Change
Where a platform runs a version for younger children, we now say so
Eleven of the guides now name the version built for under-13s and link the official page for it: YouTube Kids, Roblox Kids and Roblox Select, Fortnite's Cabined Accounts, Microsoft child accounts for Minecraft, WhatsApp's new parent-managed accounts, kids profiles on Netflix, Disney+ and Prime Video, Amazon Kids on Fire tablets, and Google Kids Space. Each one also carries the catch, because most of them have one. Netflix quietly switches off the kid-safe features if you raise the maturity rating. Prime Video downloads from an adult profile are still reachable from the Kids profile. YouTube Kids has search on by default. Where a platform has nothing for under-13s, which is Discord, Instagram, Snapchat and Twitch, the guide says that plainly instead of implying a setting exists.
Correction
The Roblox guide predated Roblox Kids and Roblox Select
Roblox introduced age-based account types in April 2026 and made them available worldwide in June. Our guide was written before that and started at Content Maturity, which is now the second decision rather than the first. There is a step ahead of it on checking which account type your child is on. For years it was true that Roblox had no child account type, and that has stopped being true.
Change
The age warning now covers under-7s, and tells you something useful
Picking an app built for 13 and over always warned you, except for the Under 7 band, where it was switched off on the assumption it would fire on every tile. It would not: six of the thirteen apps carry a 13 minimum, not all of them. The warning is back for under-7s, and where a younger version of the app exists it now offers that instead of just recording your decision.
Change
Six new guides: Fire tablet, WhatsApp, Twitch, Netflix, Disney+ and Prime Video
Amazon Fire tablets are often a child's first device and were missing entirely. Two of the new guides also correct things widely got wrong elsewhere: WhatsApp's minimum age is 13 in the UK and EU, not the 16 most parenting sites still print, and both WhatsApp and Twitch added real parental controls in 2026 that older guidance predates. Twitch's is opt-in and can be switched off by the teen, which the guide says plainly.
Change
Plans instead of one guide at a time
Setting up a child's device meant reading four separate guides and doing the same device-level step in each. Now you pick the device, the age and the apps they use, and get one plan with the overlaps merged, broken into roughly ten-minute sittings. The individual guides are unchanged and still work on their own.
Change
Weekly source checking turned on
Every Monday, each guide's official source pages are re-read and compared against what we publish. Guides confirmed unchanged get their date updated automatically. Anything that looks different is flagged for a person to check, and never rewritten automatically. The results appear here.
Correction
Crisis pages were quoting the wrong country's police
The 'something has happened' pages named specific national agencies inside advice that is identical everywhere, so a parent in the UK was reading about An Garda Síochána and a Canadian agency. The advice is now written without naming anyone's agency, and country names appear only in the reporting routes, in country-specific notes, and in one citation under a statistic. Found by a reader, not by us.
Change
Added the 'something has happened' pages
Every route through SafeStart assumed you were setting things up in advance. There was nothing for the parent whose child has already been contacted, threatened over an image, or bullied. There is now, with the reporting routes for the US, UK, Canada and Ireland, checked against each organisation's own site.
Correction
Spending advice showed two currencies at once
Recommended spending limits read '£0 / $0' because the site had no idea where you were. There is a country setting now, and the currency follows it.
Verification
All 17 guides written and verified
Every step checked against the platform's own documentation. Guides where the official source could not confirm a detail are marked with a lower source confidence on the page itself rather than presented as certain.
This page is generated from the same file the weekly check writes to, so it cannot quietly fall behind what actually happened.
