Three companies now decide whether a teenager can open a chat window on Roblox, join a voice call on Discord, or turn on cross-play on PlayStation. None of them make games. Yoti, k-ID, and Persona have quietly become the compliance backbone of the gaming industry’s biggest age-verification push in a decade, triggered by Australia’s under-16 social media law, the UK’s Online Safety Act, and a stack of new US state rules. By September 27, 2026, each vendor is embedded in a different corner of the gaming world, and the technology, accuracy, and privacy trade-offs behind each one differ more than most players realize.
This comparison breaks down how Yoti, k-ID, and Persona actually work, which platforms use which vendor, what the accuracy numbers really show, what each check costs a studio to run, and how the privacy fight between Discord and Persona reshaped the entire category. If you build, operate, or just play on a platform that now asks for a selfie or an ID scan before you can chat, this is the vendor landscape behind that prompt.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
Why Age Verification Became Non-Negotiable for Gaming Platforms in 2026
Age verification in gaming didn’t arrive because platforms wanted it. It arrived because regulators stopped accepting a birthdate typed into a signup form as proof of anything. Australia’s social media minimum age law took effect on December 10, 2025, requiring age-restricted platforms to take “reasonable steps” to prevent Australians under 16 from creating or keeping an account, according to the Office of the Australian Information Commissioner. The law doesn’t name a specific vendor, but it forced every major platform operating in Australia to stand up an age-assurance pipeline almost overnight, which is exactly why Steam’s own Australian age checks and Epic Games Store’s age verification rollout both landed within months of each other.
The UK’s Online Safety Act added a second front. Yoti says July 25, 2025 marked the point when many pornography, social-media, livestreaming, dating, and some gaming services in the UK were legally required to run age checks, a deadline documented on the company’s own CEO update from August 2025. The EU’s Digital Services Act adds pressure on very large online platforms to assess and mitigate risks to minors, without mandating a single technology. In the US, a patchwork of state laws in places like Texas, Louisiana, and Utah is pushing age checks into app stores and platforms serving minors, though the exact obligations vary by state and remain subject to ongoing litigation.
None of these laws tell a platform which vendor to use. That gap is exactly where Yoti, k-ID, and Persona compete. Each sells the same basic promise, turning a face, a document, or a data signal into a yes-or-no answer about whether a player clears an age threshold, but the technical approach, the accuracy claims, and the privacy posture behind that yes-or-no answer diverge sharply once you look past the marketing page.
How Yoti’s Facial Age Estimation Works
Yoti is the oldest and most broadly deployed of the three vendors in gaming specifically. Its age verification page for gaming customers lists six proof-of-age methods built around documents, data checks, and a selfie, and the company states plainly: “Our online portal offers your customers six easy proof of age methods, using documents, data or a selfie.” A platform sets an age threshold, and Yoti returns a binary result rather than handing over a birthdate, an approach the company frames as privacy by design: “You tell us your age threshold and we’ll give you a yes/no answer.”
Yoti’s facial age estimation runs on its own trained model line, most recently yoti-004, and the company reports it does not train the model to recognize gender, skin color, or to match one face against another, limiting it strictly to estimating an age range from a live selfie. On accuracy, Yoti cites a National Institute of Standards and Technology mean absolute error of 1.71 years for white males aged 18 to 30 within a specific image-resolution band, while its own white paper reports 1.83 years for a narrower demographic slice. A separate Reuters report on the broader age-assurance industry cites Yoti’s claimed 1.04-year average error specifically for children in the 14-to-18 range, though that figure comes from the company rather than an independent head-to-head test against rivals.
Two independent-style credentials back Yoti’s liveness detection specifically, rather than its age estimation. The company says its passive biometric liveness detection was the first to pass iBeta Level 3 presentation-attack-detection testing, the highest liveness standard under ISO/IEC 30107, according to its January 2026 company update. Yoti also holds Kantara Initiative certification as an Orchestration Service Provider under the UK’s Digital Identity and Attributes Trust Framework. Those are real third-party audits, but they test whether Yoti can detect a photo or mask being held up to a camera, not whether its age estimate is correct.
How k-ID’s Age Signal System Works
k-ID takes a different starting point. Rather than positioning itself as a single verification method, k-ID frames its product around what it calls an “age signal,” a piece of information about a player’s age that determines what permissions and features they get access to, according to k-ID’s own developer documentation. In practice, that means k-ID can lean on data a platform already has, on parental or guardian confirmation, or on additional verification challenges such as facial estimation or document checks, depending on the jurisdiction and the risk level of the feature being gated.
Discord is k-ID’s flagship gaming deployment. Discord’s own support documentation states plainly: “This is done through k-ID, a vendor we use to confirm your age on our behalf.” That single sentence is currently the clearest public confirmation of a k-ID gaming integration at scale. k-ID’s developer docs also describe an “age-gate” concept where a trustworthy platform-provided age signal can satisfy a compliance requirement without triggering a fresh verification challenge, in the company’s words: “A trustworthy platform age signal can satisfy assurance without the challenge being issued.” That design choice matters because it lets a platform reuse a signal a user already provided elsewhere, cutting down on how often someone has to hand over a selfie or ID.
The tradeoff is that k-ID’s accuracy story is far less public than Yoti’s or Persona’s. No independent test result or vendor-reported average-error figure for k-ID’s own facial estimation model turned up in current public reporting, which makes a direct numerical accuracy comparison against Yoti or Persona impossible to responsibly state today. What’s verifiable is adoption and design philosophy, not a benchmark score.
How Persona’s Identity Verification Works
Persona is the vendor behind Roblox’s age-verification system, and Roblox is by far the largest gaming platform running any of these three services. Roblox’s support documentation describes two paths through Persona: a facial-age-estimation flow using an image or short video, and a full identity-verification flow using a government-issued passport, driver’s license, state ID, or residence permit for players who want expanded social features or age-appropriate content access.
Roblox is specific about what happens to that data. Its official newsroom post states: “Images and videos completed through Facial Age Estimation are securely processed by our vendor, Persona, and deleted immediately after processing.” On the document-based path, Roblox says it does not receive the raw identification information submitted to Persona, keeping the sensitive document data inside Persona’s systems rather than Roblox’s own. Reddit is the other major platform tied to Persona, referenced by Discord’s own blog as “a company used by platforms like Roblox and Reddit,” though Reddit has not published detailed technical specifics of its own implementation.
On accuracy, Reuters reports Persona advertises an average age-estimation error of 1.77 years for players aged 13 to 17. Set next to Yoti’s cited 1.04-year figure for a similar teen age band, that’s a real, sourced gap of roughly 0.73 years, but it comes with a major caveat worth repeating: these are two different companies’ self-reported numbers, measured on different datasets under different conditions, not a controlled side-by-side test. Treat it as a starting point for questions to ask a vendor during procurement, not as a settled verdict on which system is more accurate in the field.
Yoti vs k-ID vs Persona: Full Specifications Compared
The table below lines up what’s publicly documented about each vendor’s technology, certifications, and gaming footprint as of September 2026. Fields marked “not publicly disclosed” reflect genuine gaps in public reporting rather than a company choosing to hide the number from this article.
| Category | Yoti | k-ID | Persona |
|---|---|---|---|
| Primary verification methods | Facial age estimation, ID document, data checks, digital ID | Platform age signals, parental/guardian confirmation, facial estimation, document checks | Facial age estimation, government ID (passport, license, state ID, residence permit) |
| Core age-estimation model | yoti-004 (successor to yoti-003) | Not publicly disclosed as a named model | Persona’s proprietary facial-estimation model |
| Vendor-reported teen accuracy | 1.04-year average error, ages 14-18 (per Reuters) | Not publicly disclosed | 1.77-year average error, ages 13-17 (per Reuters) |
| Independent testing referenced | NIST mean-absolute-error testing on yoti-004 (1.71-1.83 years, narrow demographic) | Not publicly disclosed | Not publicly disclosed independent of vendor claim |
| Liveness / anti-spoof certification | iBeta Level 3 presentation-attack detection (ISO/IEC 30107) | Not publicly disclosed | Not publicly disclosed |
| On-device processing option | Not confirmed as a standard offering | Framework supports platform-held signals to reduce re-verification | Cited by Discord as not fully on-device during its UK test |
| Data retention on facial checks | Data-minimization architecture; retention varies by integration | Not publicly disclosed | Deleted immediately after processing (per Roblox) |
| Compliance frameworks cited | UK DIATF (Kantara-certified Orchestration Service Provider) | ISO/IEC 27566-1 certified via ACCS; age-signal orchestration framework | Enterprise identity-verification compliance stack |
| Confirmed gaming/social clients | PlayStation, Bluesky (via Epic Games arrangement) | Discord | Roblox, Reddit |
| Integration model | Online portal plus API, six-method selector | Developer-facing age-signal and age-gate APIs | Enterprise identity-verification API |
| Pricing model | Quotation-based enterprise pricing | Quotation-based enterprise pricing | Quotation-based enterprise pricing |
| Notable 2025-2026 development | iBeta Level 3 liveness certification; Epic/Bluesky deployment | Became Discord’s sole age-assurance vendor after the Persona test | Lost Discord’s business over on-device processing requirement |
Accuracy Benchmarks: NIST Scores and Vendor-Reported Error Rates
Every accuracy number circulating about these three vendors deserves a health warning, because none of them come from the same test. Yoti points to National Institute of Standards and Technology testing to support its claim that its yoti-003 model ranked first for estimating ages 13 to 16, and the company’s own white paper reports a 1.83-year mean absolute error for a specific demographic slice under yoti-004, versus a 1.71-year NIST figure for white males aged 18 to 30 under a defined image-resolution range. Persona’s 1.77-year figure for ages 13 to 17, as reported by Reuters, comes from the company’s own marketing rather than a NIST evaluation. k-ID has not published a comparable error figure at all in current reporting.
Reading these numbers side by side without the caveats would suggest Yoti is meaningfully more accurate than Persona at estimating a teenager’s age, and it might be. But age-estimation error rates shift by sex, skin tone, image quality, lighting, and the specific age band being tested, so a number generated on one demographic slice doesn’t automatically generalize to a platform’s full, diverse user base. No public report so far has run Yoti, k-ID, and Persona through an identical dataset with the same demographic breakdown and the same false-positive and false-negative reporting. Until that head-to-head exists, “which vendor is most accurate” remains a question with a partial, vendor-supplied answer rather than a definitive one.
Pricing Comparison: What Age Checks Cost Game Studios
None of the three vendors publishes a standard price sheet, which is typical for enterprise identity infrastructure sold on volume and integration complexity. What is public is an industry-wide cost benchmark reported by Reuters, based on interviews with age-assurance industry executives: basic, machine-only age-assurance checks generally run well under $1 per check, with high-volume enterprise deals sometimes falling to single-digit cents per check. That’s an industry average, not a confirmed line-item price for Yoti, k-ID, or Persona specifically, but it’s the most concrete public cost anchor available for budgeting a rollout.
| Vendor / benchmark | Pricing model | Reported cost range | Contract type |
|---|---|---|---|
| Yoti | Quotation-based, per-check or platform license | Not publicly disclosed | Enterprise, negotiated |
| k-ID | Quotation-based, platform integration | Not publicly disclosed | Enterprise, negotiated |
| Persona | Quotation-based, per-verification API pricing | Not publicly disclosed | Enterprise, negotiated |
| Industry benchmark (low-volume) | Machine-only automated check | Well under $1 per check | Standard API usage |
| Industry benchmark (high-volume) | Machine-only automated check | Single-digit cents per check | Large-scale enterprise deal |
For context, a platform the size of Roblox processing age checks across a meaningful fraction of its user base is operating at a scale where even a fraction of a cent per check moves the total cost by millions of dollars a year, which is exactly why volume-based enterprise contracts, not published price lists, dominate this market.
Which Gaming Platforms Use Which Vendor: Real-World Deployments
Public confirmation of which platform uses which vendor is patchier than you’d expect for infrastructure this consequential. Here’s what’s actually documented, platform by platform.
| Platform | Vendor | Context |
|---|---|---|
| Roblox | Persona | Facial age estimation and government-ID verification for chat access and age-appropriate content, expanded worldwide per Roblox’s newsroom announcement |
| Discord | k-ID | Sole confirmed age-assurance vendor per Discord’s own support documentation, after a rejected Persona trial |
| PlayStation (UK) | Yoti | Mobile number, facial scan, or ID verification options for voice chat access, per PlayStation’s UK support FAQ |
| Bluesky | Yoti (via Epic Games) | Epic Games supplies Yoti’s facial age estimation to Bluesky for UK users, per Yoti’s press archive |
| Persona | Referenced by Discord’s blog as a Persona client alongside Roblox; Reddit has not detailed its own implementation | |
| Steam | No confirmed vendor | Runs its own regional age-check flow (including credit-card verification in Australia) without a publicly confirmed Yoti, k-ID, or Persona integration |
| Xbox | No confirmed vendor | Rolled out its own age-verification deadline without a publicly named third-party vendor in current reporting |
The pattern that emerges is a platform-by-platform bet rather than an industry standard. Roblox and Reddit lean on Persona’s broader identity-verification stack. Discord bet on k-ID’s platform-signal model after testing and rejecting Persona. Sony and Epic both went with Yoti, likely drawn by its certified liveness detection and its multi-industry track record outside gaming, including retail age checks for alcohol sales through partners in the UK and Channel Islands. Steam and Xbox, meanwhile, appear to be running their own in-house or region-specific flows rather than outsourcing to any of these three, which is worth remembering the next time Xbox and PlayStation age-check deadlines get compared as if they were built the same way.
Privacy and Biometric Data Handling Compared
Every one of these vendors processes something sensitive, whether that’s a face, a government ID, or behavioral data used to infer an age band, and the differences in how long that data sticks around matter as much as the accuracy numbers. Yoti’s public positioning centers on data minimization: the company says its architecture is built so it knows very little about the people using its service, and its facial-estimation model is explicitly not trained to identify who someone is, only to estimate an age range. Persona’s specific commitment, as stated by Roblox, is that images and video used for facial age estimation are deleted immediately after processing, and that raw government-ID data submitted through Persona’s document-verification flow never reaches Roblox’s own systems.
k-ID’s public privacy documentation is thinner on specifics. Its developer materials describe the age-signal and age-gate concepts in detail. On certification, k-ID holds certification under ISO/IEC 27566-1, the international standard for age-assurance systems, assessed through the Age Check Certification Scheme (ACCS), and its AgeKey reusable credential used on Discord is certified under the same standard. Because k-ID operates as an orchestration layer across multiple underlying age-assurance providers rather than a single verification method, the accuracy of any given check belongs to the provider it routes to, and those providers publish their own figures. k-ID’s data-retention terms are set out in section 9 of its privacy policy.
None of the three vendors has a publicly reported data breach tied to their age-verification product as of this writing. The absence of a disclosed incident isn’t proof one hasn’t happened, but it’s worth noting given how frequently biometric and identity data breaches make headlines elsewhere in the industry, from gaming account takeovers to large-scale identity-document leaks at other verification providers.
The Regulatory Wave Driving Adoption: Australia, UK, EU, and US
Australia’s under-16 social media law is the sharpest deadline in this story. From December 10, 2025, age-restricted platforms operating in Australia must take reasonable steps to prevent under-16 users from creating or keeping accounts, according to the OAIC’s official guidance. The law is deliberately technology-neutral, which is exactly why Steam, Epic, Xbox, and PlayStation have each built or bought slightly different compliance flows rather than converging on one shared vendor.
Complying without alienating your user base
The UK’s Online Safety Act adds a second, earlier deadline. Yoti dates the enforcement trigger to July 25, 2025, covering porn, social media, livestreaming, dating, and some gaming services operating in the UK. The EU’s Digital Services Act layers on a third obligation, pressuring very large online platforms to assess and mitigate risks to minors without naming a specific verification method, leaving room for platforms to choose facial estimation, document checks, or parental-consent flows depending on their risk tolerance and user base. In the US, state-level laws in Texas, Louisiana, and Utah are pushing age verification into app stores and platforms serving minors, though the scope and enforcement status of each law varies and several remain tied up in litigation.
The net effect is a compliance patchwork rather than a single global standard, and that patchwork is precisely why a platform like Roblox, which operates in more than 180 countries, needs a vendor like Persona that can flex between a lightweight facial estimate in one jurisdiction and a full government-ID check in another, rather than a one-size-fits-all product.
Why Discord Rejected Persona: The On-Device Privacy Fight
The clearest public case study in vendor selection anywhere in this space comes from Discord itself. In a company blog post titled “Getting Global Age Assurance Right,” Discord disclosed that it ran a limited test with Persona in the UK in January 2026. The company evaluated Persona specifically because, as Discord noted, it’s “a company used by platforms like Roblox and Reddit.” After completing the test, Discord decided not to move forward, and said all data collected during the trial was deleted in line with its privacy policy.
The stated reason wasn’t accuracy. It was architecture. Discord said its requirement was for facial age estimation to happen entirely on-device, so biometric data never leaves a user’s phone, and that Persona’s approach at the time didn’t meet that bar. That’s a meaningfully different failure mode than a vendor being rejected for being wrong about someone’s age; it’s a rejection over where the processing physically happens and how much data ever has to leave a user’s device in the first place.
This episode is the single most concrete public signal of how differently these three companies approach the privacy side of age assurance. k-ID’s design, at least as far as it ended up deployed at Discord, aligns with an on-device-first philosophy. Persona’s architecture, as tested by Discord, processes more centrally. Yoti’s public materials emphasize data minimization but don’t make an explicit on-device processing claim comparable to what Discord says it required. If your organization treats on-device biometric processing as a hard requirement rather than a nice-to-have, Discord’s public writeup is the closest thing to an independent field test any of these three vendors has received.
Migration Guide: How Game Studios Choose and Implement a Vendor
Studios rolling out age verification for the first time, or replacing an ad-hoc birthdate field with a real vendor, tend to go through the same sequence of decisions regardless of which of the three companies they end up choosing.
- Map regulatory exposure by region first. A studio operating in Australia, the UK, the EU, and specific US states is subject to four different legal definitions of “reasonable steps,” and the vendor needs to support the weakest-common-denominator flow for each region rather than a single global default.
- Decide whether on-device processing is a hard requirement. Discord’s public rejection of Persona over this exact issue shows it’s a legitimate deal-breaker, not a theoretical concern. Get a specific, written answer from any vendor about where facial data is processed before signing anything.
- Evaluate the fallback methods, not just the primary one. Every vendor here supports more than one verification path. A platform that only offers a facial scan will lose users who can’t or won’t provide one; check whether the vendor’s ID-document and data-check fallbacks are integrated as smoothly as the facial flow.
- Get the data-retention policy in writing. Persona’s “deleted immediately after processing” commitment to Roblox is the clearest public retention statement among the three; ask k-ID and Yoti for an equivalent written commitment specific to your integration rather than relying on general marketing language.
- Negotiate pricing against the Reuters-reported industry benchmark. Well under $1 per check at low volume, single-digit cents at high volume, gives a studio a concrete anchor for vendor negotiations rather than accepting a first quote at face value.
- Plan an appeals path for false results. Age-estimation error rates of roughly 1 to 2 years mean some 17-year-olds will be flagged as under a threshold and some 15-year-olds will slip through; build a manual review or document-upload escalation path before launch, not after user complaints arrive.
- Stage the rollout by region. Most platforms covered here, including Discord’s own age-check rollout, launched region by region rather than globally on day one, which limits the blast radius if the integration needs adjustment.
A simplified version of the decision logic a platform builds around any of these vendors looks like this, illustrating the age-gate concept rather than any specific vendor’s real source code:
// Illustrative age-gate flow, not vendor source code
function checkAgeGate(user, platformAgeSignal, ageVerificationVendor) {
if (platformAgeSignal && platformAgeSignal.trustworthy) {
return platformAgeSignal.ageBand; // reuse an existing trusted signal
}
const result = ageVerificationVendor.estimate({
method: "facial_estimation", // or "id_document", "data_check"
threshold: platform.minimumAge
});
return result.passed ? "granted" : "restricted";
}
Pros and Cons of Yoti, k-ID, and Persona
No vendor here is a clean win across every category. Each one trades off differently between accuracy documentation, privacy architecture, and gaming-specific footprint.
- Yoti pros: Certified iBeta Level 3 liveness detection, published NIST-referenced accuracy data, six verification methods, multi-industry track record beyond gaming, and confirmed deployments at PlayStation and Bluesky.
- Yoti cons: No explicit on-device processing guarantee comparable to what Discord required, pricing entirely quotation-based with no public benchmark specific to the company, and its strongest accuracy figures are drawn from narrow demographic test slices.
- k-ID pros: Purpose-built “age signal” orchestration that routes across multiple age-assurance providers to minimize repeated verification friction, certification under ISO/IEC 27566-1 (via ACCS) with its AgeKey credential certified under the same standard, and sole vendor at Discord after a competitive evaluation against Persona.
- k-ID cons: As an orchestration layer it does not publish a single headline accuracy figure of its own (those belong to the underlying providers it routes to), and it has a single confirmed major gaming client so far.
- Persona pros: The largest raw user base of any vendor here through Roblox, a clear and specific data-deletion commitment for facial checks, and a dual facial-estimation-plus-document-verification model that covers more jurisdictions.
- Persona cons: Lost a competitive evaluation at Discord specifically over its processing architecture, a higher vendor-reported average error (1.77 years) than Yoti’s cited figure for a similar age band, and less liveness-specific third-party certification than Yoti.
What the Vendors Say
Each company’s own public documentation reveals as much about its design philosophy as any third-party analysis. k-ID’s developer documentation defines its core product this way:
“An age signal is information about a player’s age that k-ID uses to determine what permissions and features they should have access to.”
k-ID, Developer Documentation (docs.k-id.com)
That same documentation explains why k-ID doesn’t always need to run a fresh check:
“A trustworthy platform age signal can satisfy assurance without the challenge being issued.”
k-ID, Developer Documentation (docs.k-id.com)
Yoti’s gaming page frames its own product around simplicity for the integrating platform rather than the underlying technical complexity:
“Our online portal offers your customers six easy proof of age methods, using documents, data or a selfie.”
Yoti, official product page (yoti.com/gaming)
These statements aren’t neutral analysis, they’re each company describing its own product. But read together, they map cleanly onto the real-world split documented earlier: k-ID optimizing for signal reuse and reduced friction, and Yoti optimizing for method breadth and a simple integration surface.
Real-World Examples: Five Deployments That Show the Split
Reading the vendor comparisons in the abstract only goes so far. Five live deployments show how differently Yoti, k-ID, and Persona actually behave once they’re wired into a real platform with millions of concurrent users.
- Roblox worldwide chat gate (Persona). Roblox’s global chat age-check rollout, announced through the company’s own newsroom, expanded Persona-powered facial estimation and ID verification to Roblox’s entire worldwide user base to gate chat access and age-appropriate content. Given Roblox’s scale, this is almost certainly the single largest gaming deployment of any of these three vendors, processing checks across more than 180 countries with two distinct verification paths running in parallel.
- Discord’s platform-wide age assurance (k-ID). Discord’s shift to k-ID followed a documented, public evaluation-and-rejection cycle involving a competing vendor, making it the clearest evidence-based vendor-selection story anywhere in this category. Discord’s own support documentation now names k-ID directly as the company confirming user ages on its behalf, a level of transparency none of the other platforms in this comparison has matched.
- PlayStation UK voice chat access (Yoti). Sony’s UK rollout through Yoti offers three distinct verification paths: mobile number, facial scan, or ID document. That three-way fallback structure gives PlayStation users more flexibility than a single-method system would allow, and it mirrors the same multi-method philosophy Yoti advertises on its dedicated gaming product page.
- Bluesky’s UK age checks (Yoti, via Epic Games). Bluesky’s age checks for UK users run on Yoti’s facial-age-estimation technology, but they arrive through an unusual channel: supplied by Epic Games rather than built or bought directly by Bluesky itself. That arrangement shows how age-verification infrastructure is starting to move between companies as a shared utility rather than each platform building or buying its own stack from scratch.
- Reddit’s identity checks (Persona). Reddit’s use of Persona is confirmed only through a passing mention in Discord’s own blog post, which described Persona as “a company used by platforms like Roblox and Reddit,” rather than through a direct Reddit announcement. That secondhand confirmation is itself a useful data point: it shows how much of this vendor landscape is still being pieced together from adjacent company statements rather than clean, first-party disclosures from every platform involved.
Two platforms conspicuously absent from this vendor map are Steam and Xbox, both of which appear to be running their own region-specific verification logic rather than outsourcing to Yoti, k-ID, or Persona. That’s a meaningful data point for anyone assuming every major platform has converged on the same handful of third-party providers: at least two of the industry’s largest players have chosen to build rather than buy.
Verdict: Which Age Verification Vendor Wins in 2026
There’s no single winner here, because “winning” depends on what a studio actually optimizes for. If certified liveness detection and a documented accuracy trail matter most, Yoti currently has the strongest public paper trail, backed by iBeta Level 3 certification and NIST-referenced testing that neither k-ID nor Persona can currently match in public reporting. If minimizing repeated friction for returning users and aligning with an on-device-first privacy stance matters most, k-ID’s design and its selection by Discord after a competitive evaluation make it the safer default, even though its own accuracy numbers remain unpublished. If raw scale and a proven track record processing identity documents across 180-plus countries matters most, Persona’s position inside Roblox is unmatched by either competitor, even after its loss at Discord.
Studio size should factor into the decision as much as the technology itself. A large platform like Roblox or Discord can negotiate custom enterprise terms, run its own privacy audits, and demand a written on-device processing guarantee before signing. A smaller studio bolting age verification onto an existing game for the first time typically has far less negotiating leverage, and will likely find itself accepting whichever vendor offers the fastest SDK integration and the clearest self-serve documentation, which today points toward Yoti’s six-method portal as the least friction-heavy starting point for a first-time integration. That’s a practical consideration, not an endorsement, and it can change quickly as k-ID and Persona both continue publishing more integration tooling through 2026.
The one number that should worry every studio evaluating these vendors isn’t a competitive comparison at all: it’s the roughly 1-to-2-year error margin that all three companies’ facial-estimation approaches carry near the 13-to-18 age boundary that regulators actually care about. That margin is small enough to sound impressive in a press release and large enough to misclassify real teenagers on either side of a legal cutoff. Whichever vendor a studio picks, the honest verdict is that facial age estimation alone isn’t precise enough to be the only method offered, which is exactly why every vendor in this comparison, and every platform using them, still keeps a document-based or data-based fallback in place.
Frequently Asked Questions
What is age verification technology in gaming, and why did it suddenly become mandatory in 2026?
Age verification technology estimates or confirms a user’s age through a selfie, an ID document, or platform data, rather than relying on a self-reported birthdate. It became mandatory across major gaming and social platforms because of overlapping laws, including Australia’s under-16 social media law effective December 10, 2025, the UK’s Online Safety Act, and the EU’s Digital Services Act, all of which put legal liability on platforms rather than users.
Is Yoti, k-ID, or Persona more accurate at estimating a player’s age?
Based on vendor-reported figures cited by Reuters, Yoti claims a 1.04-year average error for ages 14-18, while Persona claims a 1.77-year average error for ages 13-17. k-ID has not published a comparable figure. These numbers come from different tests under different conditions, not a single independent head-to-head comparison, so they should be treated as a starting point for questions rather than a final ranking.
Does Roblox, Discord, or PlayStation store my face scan or ID after an age check?
Roblox states that images and video used for Persona’s facial age estimation are deleted immediately after processing, and that Roblox itself doesn’t receive raw ID document data submitted through Persona. Yoti describes a data-minimization architecture for its checks. Public documentation on k-ID’s exact retention window for Discord’s age checks is less detailed.
Why did Discord test Persona and then choose k-ID instead?
Discord publicly disclosed running a limited UK test with Persona in January 2026 and decided not to proceed, stating that its requirement was for facial age estimation to happen entirely on-device so biometric data never leaves a user’s phone, a bar it said Persona did not meet at the time. Discord’s current, sole confirmed age-assurance vendor is k-ID.
Do Steam and Xbox use any of these three vendors?
No confirmed public integration with Yoti, k-ID, or Persona exists for Steam or Xbox as of current reporting. Both platforms appear to run their own region-specific age-check processes, including Steam’s credit-card-based verification flow in Australia.
How much does it cost a game studio to add age verification?
None of the three vendors publishes a standard price list. Reuters reports an industry-wide benchmark of well under $1 per check for basic machine-only age assurance at low volume, dropping to single-digit cents per check at high enterprise volume, though actual contract terms are negotiated individually.
Can players avoid the facial scan and use another verification method instead?
Generally yes. Yoti offers six proof-of-age methods including documents and data checks alongside its facial scan. Persona offers a government-ID path as an alternative to facial estimation on Roblox. k-ID’s age-signal model can also satisfy verification without a fresh facial or document check if a trustworthy signal already exists from the platform.
What happens if an age-estimation system gets my age wrong?
All three vendors’ facial-estimation approaches carry a real margin of error, roughly 1 to 2 years based on the vendor-reported figures for Yoti and Persona. Because of this, platforms using these vendors typically keep a document-upload or manual-review fallback in place for users who are incorrectly flagged, rather than relying on the facial estimate as a final, unappealable decision.
Related Coverage
- Salt vs Wallarm vs Akamai API Security: 48x G2 Gap [2026]
- Norton 360 for Gamers vs Bitdefender vs Malwarebytes: 3% FPS Gap [2026]
- Reality Defender vs Hive vs Sensity: $33M vs $3.2M Gap [2026]
- AppOmni vs Obsidian vs CrowdStrike: 100-App SSPM Gap [2026]
- ShinyHunters Bypass WAF on CVSS 9.8 PeopleSoft Bug [2026]
