Immersive
June 27, 2025
What Are Spatial Anchors and Why They Matter
Breaking Down Spatial Anchors in AR

Breaking Down Spatial Anchors in AR/MR

Augmented Reality (AR) and Mixed Reality (MR) depend on accurate understanding of the physical environment to create realistic experiences, and they hit this target with the concept of spatial anchors. These anchors act like markers, either geometric or based on features, that help virtual objects stay in the same spot in the real world — even when users move around.

Sounds simple, but the way spatial anchors are implemented varies a lot depending on the platform; for example, Apple’s ARKit, Google’s ARCore, and Microsoft’s Azure Spatial Anchors (ASA) all approach them differently.

If you want to know how these anchors are used in practical scenarios or what challenges developers often face when working with them, this article dives into these insights too.

What Are Spatial Anchors and Why They Matter

A spatial anchor is like a marker in the real world, tied to a specific point or group of features. Once you create one, it allows for some important capabilities:

  1. Persistence. Virtual objects stay exactly where you placed them in the real-world, even if you close and restart the app.
  2. Multi-user synchronization. Multiple devices can share the same anchor, so everyone sees virtual objects aligned to the same physical space.
  3. Cross-session continuity. You can leave a space and come back later, and all the virtual elements will still be in the right place.

In AR/MR, your device builds a point cloud or feature map by using the camera and built-in sensors like the IMU (inertial measurement unit). Spatial anchors are then tied to those features, and without them, virtual objects can drift or float around as you move, shattering the sense of immersion.

Technical Mechanics of Spatial Anchors

At a high level, creating and using spatial anchors involves a series of steps:

Feature Detection & Mapping

To start, the device needs to understand its surroundings: it scans the environment to identify stable visual features (e.g., corners, edges). Over time, these features are triangulated, forming a sparse map or mesh of the space. This feature map is what the system relies on to anchor virtual objects.

Anchor Creation

Next, anchors are placed at specific 3D locations in the environment in two possible ways:

  • Hit-testing. The system casts a virtual ray from a camera to a user-tapped point, then drops an anchor on the detected surface.
  • Manual placement. Sometimes, developers need precise control, so they manually specify the exact location of an anchor using known coordinates, like ensuring it perfectly fits on the floor or another predefined plane.

Persistence & Serialization

Anchors aren’t temporary — they can persist, and here’s how systems make that possible:

  • Locally stored anchors. Frameworks save the anchor’s data, like feature descriptors and transforms, in a package called a “world map” or “anchor payload”.
  • Cloud-based anchors. Cloud services like Azure Spatial Anchors (ASA) upload this anchor data to a remote server to let the same anchor be accessed across multiple devices.

Synchronization & Restoration

When you’re reopening the app or accessing the anchor on a different device, the system uses the saved data to restore the anchor’s location. It compares stored feature descriptors to what the camera sees in real time, and if there’s a good enough match, the system confidently snaps the anchor into position, and your virtual content shows up right where it’s supposed to.

However, using spatial anchors isn’t perfect, like using any other technology, and there are some tricky issues to figure out:

  • Low latency. Matching saved data to real-time visuals has to be quick; otherwise, the user experience feels clunky.
  • Robustness in feature-scarce environments. Blank walls or textureless areas don’t give the system much to work with and make tracking tougher.
  • Scale drift. Little errors in the system’s tracking add up over time to big discrepancies.

When everything falls into place and the challenges are handled well, spatial anchors make augmented and virtual reality experiences feel seamless and truly real.

ARKit’s Spatial Anchors (Apple)

Apple’s ARKit, rolled out with iOS 11, brought powerful features to developers working on AR apps, and one of them is spatial anchoring, which allows virtual objects to stay fixed in the real world as if they belong there. To do this, ARKit provides two main APIs that developers rely on to achieve anchor-based persistence.

ARAnchor & ARPlaneAnchor

The simplest kind of anchor in ARKit is the ARAnchor, which represents a single 3D point in the real-world environment and acts as a kind of “pin” in space that ARKit can track. Building on this, ARPlaneAnchor identifies flat surfaces like tables, floors, and walls, allowing developers to tie virtual objects to these surfaces.

ARWorldMap

ARWorldMap makes ARKit robust for persistence and acts as a snapshot of the environment being tracked by ARKit. It captures the current session, including all detected anchors and their surrounding feature points, into a compact file.

There are a few constraints developers need to keep in mind:

  • World maps are iOS-only, which means they cannot be shared directly with Android.
  • There must be enough overlapping features between the saved environment and the current physical space, and textured structures are especially valuable for this, as they help ARKit identify key points for alignment.
  • Large world maps, especially those with many anchors or detailed environments, can be slow to serialize and deserialize, causing higher application latency when loading or saving.

ARKit anchors are ideal for single-user persistence, but sharing AR experiences across multiple devices poses additional issues, and developers often employ custom server logic (uploading ARWorldMap data to a backend), enabling users to download and use the same map.

However, this approach comes with caveats: it requires extra development work and doesn’t offer native support for sharing across platforms like iOS and Android.

ARCore’s Spatial Anchors (Google)

Google’s ARCore is a solid toolkit for building AR apps, and one of its best features is how it handles spatial anchors:

Anchors & Hit-Testing

ARCore offers two ways to create anchors. You can use Session.createAnchor(Pose) if you already know the anchor’s position, or you can use HitResult.createAnchor() if you want to define the anchor’s location based on the surface detected by the system. Once created, the position of an anchor is represented as a 3×4 transform matrix, giving you control over its placement.

Cloud Anchors (Alpha → Beta → Stable)

The Cloud Anchor API allows you to save your anchors to Google’s servers and share them across devices — including iOS clients using ARCore’s SDK for iOS.

There are a few catches to be aware of:

  • Free-tier hosted anchors live for 24 hours. If you need them to last longer, you’ll need the Cloud Anchor API for business, which extends retention up to 365 days with quotas.
  • Hosting an anchor works best when the environment has lots of unique features (like textured surfaces) and good lighting, so avoid plain or dimly lit spaces, as they make data capture harder, messing up the accuracy.
  • If you’re hosting a lot of anchors in a small area, ARCore will take longer to resolve them because the system has more data to sift through.

ARCore’s Cloud Anchors enable cross-platform (Android ↔ iOS) multi-user experiences, giving an edge over ARKit, which is iOS-only. However, you’ll need to plan for API quotas and deal with the default retention limits if you’re not on the business tier.

Azure Spatial Anchors (Microsoft)

Microsoft’s Azure Spatial Anchors (ASA) is a cloud service designed to work across HoloLens, iOS, and Android, so you can create AR experiences without stressing over platform-specific details.

Cross-Platform SDK

ASA gives you a single SDK that’s compatible with Unity, Unreal, and native platforms (UWP, iOS, Android).

Anchor Persistence & Retrieval

When you create an anchor, ASA captures a 3D point cloud of its surroundings and securely uploads this data to Azure’s globally distributed backbone. What do you get? A unique 32-character unique Anchor ID that acts like a permanent address for your anchor in the cloud. And these anchors don’t have an expiration date, you decide when they’re no longer needed — they persist until you explicitly delete them.

Spatial Anchors CRUD

There are a few areas where Azure Spatial Anchors are great:

  • Cross-platform consistency. ASA enables anchors to function across different ecosystems, including iOS, Android, HoloLens, and Magic Leap apps.
  • Long-term persistence. Anchors don’t just disappear on you like they do with some other solutions. The default Azure account comes with a quota for 1,000 anchors, which should cover most projects. If not, you can scale up as needed.
  • Scale & security. Since ASA is built into Azure’s ecosystem, enterprises can manage anchor data alongside other cloud resources, integrate with Azure Active Directory, and enforce role-based access.
  • Environment requirements. One heads-up, though. ASA works best in textured environments. Large open spaces with few features (e.g., empty rooms) often fail to produce reliable anchors. But honestly, that’s the case with most AR tools now.

Azure Spatial Anchors is great for anyone building AR apps: it saves developers from dealing with all the platform-specific issues due to cross-platform support and long-term anchor persistence. Just make sure your environment has enough texture for the best results.

Real-World Use Cases

AR is being used in seriously impactful ways, and it’s wild to see some of the ways it’s being put to work. Check these out:

Multiplayer AR Games

The multiplayer app “Just a Line” by Google uses ARCore Cloud Anchors to allow users to draw in the air and see each other’s drawings as if they share the same canvas.

Meanwhile, Moth + Flame (Scavenger AR) uses ASA to let multiple players discover virtual items anchored to real-world locations.Players use their phones to find and collect these items, precisely placed using GPS.

Remote Assistance & Collaboration

AR isn’t just for fun — it also solves serious problems. For example, ThyssenKrupp Elevator Service integrates HoloLens and ASA so remote experts can mark up a machine with virtual notes and arrows, which the technician on-site sees through their AR headset. The instructions stay locked to the specific parts of the equipment, making troubleshooting faster and cutting maintenance time by around 30%.

Industrial AR Navigation

Honeywell’s Connected Plant has workers using AR glasses that project virtual arrows onto the warehouse floor, guiding them along optimized paths to pick orders faster. And this saves tons of time on order picking — around 25% faster, actually. What’s even better, anchors ensure virtual arrows stay accurate shift after shift, so the system is always reliable.

Retail & Showroom Experiences

And, of course, we can’t skip how AR brings retail experiences closer to home: IKEA Place iOS app is a standout here. Using ARKit’s local anchors, customers can place virtual furniture in their rooms to see how it’ll look and fit (via hit-testing on detected planes) and save their room setups thanks to exporting and importing ARWorldMap data.

Limitations & Common Pitfalls

Even as the tools for spatial anchors improve, they have challenges as well.

Feature Scarcity in Environments

Some places just don’t provide enough visual details for anchors to work well — for example, empty white walls, uniform floors, or large glass areas with little texture. Anchors may fail to be created or matched reliably.

Dynamic environments add to the challenge: moving objects (people, equipment) can occlude reference features, leading to tracking issues.

Lighting Variations

Lighting matters more than you think: abrupt changes, like turning the lights off or moving to a darker area, can mess with how anchors are tracked, they may “jump” or even temporarily disappear as the system struggles to adjust.

Scale & Drift

Small tracking errors can pile up over time. The “drift” means virtual objects don’t stay exactly where they should. In this case, anchors recalibrate positioning, but virtual content can slowly diverge from intended positions without regular anchor updates.

Cross-Platform Discrepancies

Devices differently handle some basics. For example, iOS uses one type of coordinate system and Android another. While ASA translates between the two, developers still need to be careful when working with raw data.

Accuracy also varies: ARKit anchors may be more precise in small, highly textured rooms, whereas cloud anchors (ARCore/ASA) may take longer to resolve in feature-poor spaces.

Networking & Quotas

Cloud anchors rely on a good internet connection, and hosting or resolving anchors fails without it.

Free options like ARCore’s also come with limits: for instance, the free tier only keeps cloud anchors alive for 24 hours. If you’re working on large-scale projects and blow past your quota, everything will slow down or stop altogether unless you’ve set up a proper Azure SKU.

Best Practices & Recommendations

These tips will help make your spatial anchors work like a charm.

Environment Scanning

Instruct users to move slowly and sweep the device camera across all surfaces. The more details the system sees, like furniture, paintings, or posters, the better the anchors will perform. If there are plain walls or empty spaces, add some textures to your scanning route, and you’ll thank yourself later.

Anchor Density & Management

Don’t create anchors in one spot, create a mesh hierarchy instead. Start with a main anchor, then add secondary anchors for fine detail. If you’re not using anchors anymore, get rid of them to stay within service quotas (ASA) and reduce locate times.

Error Handling & Recovery

If an anchor suddenly stops tracking or becomes “limited”, show users a clear message, like “Re-scanning environment to find anchor…” to guide them. Re-scan the area regularly if you notice objects drifting out of place (for example, by more than 0.2 meters). If things still feel off, you should create a fresh anchor.

Cross-Platform Testing

Always test your setup on the devices you plan to use in real-world settings, like offices with fluorescent lights or spaces with natural daylight and some clutter.

Check how long it takes for an anchor to start working after launching. If it takes longer than five seconds to stabilize, it’s time to tweak it for better user experience.

Conclusion

Spatial anchors keep AR and MR experiences grounded, synchronized, and useful, they’re essential for persistence and multi-user synchronization. Each platform brings unique strengths:

  • ARKit (iOS-only) does fast local mapping with ARWorldMap, great for single-user setups.
  • ARCore (Android/iOS) lets you share across platforms with Cloud Anchors, though the free version only keeps them active briefly.
  • Azure Spatial Anchors offers long-term, reliable syncing and cross-platform support, ideal for big, professional setups.

To get the most out of spatial anchors, you need to understand how they work. Know how they map spaces, where they’re strong, and where they’ll give you trouble. You’ll get better results by scanning environments thoughtfully, not overloading an area with anchors, and testing under real-world conditions on all your devices.

Latest Articles

Whitepaper
August 24, 2026
Scaling Machinery Sales with Product Twins: Solutions for Digital 3D Presentations

B2B machinery buying has changed, driven by a new generation of digital-first stakeholders (Millennials and Gen Z now represent 71% of B2B buyers). They reach out with expectations already formed, alternatives compared, and, in most cases, a preferred supplier identified. For manufacturers of large, complex equipment, that leaves a narrow window to influence the decision, and static PDFs, renders and videos rarely fill it. The fix starts with something most manufacturers already have. If you hold a 3D model of your equipment, it can become an interactive presentation asset: shown at full scale in VR, explored in a browser, or configured live on a tablet in a meeting. Up to 85% lower cost per event Our new whitepaper compares traditional sales methods (trade shows, physical demos, and international product presentations) with a VR-based approach and puts concrete numbers on the difference: Trade show presence: ~€340K → €30–50K per event. A year of international demos: ~€200K → €40K. Three scenarios, one model The whitepaper also breaks down three places an immersive 3D model does the work, each with its own logic: Sales presentations: Guide a prospect through a full-scale, interactive machine. Trade fairs: Show the whole catalog from a 50 m² booth instead of shipping one or two machines to a 200 m² stand. Inbound lead generation: Let buyers explore the machine in the browser before they fill out a form. It also includes real deployments from construction, mining, elevator, and heat transfer manufacturers, plus the buyer-behavior data behind the shift to digital-first sales. If you sell equipment that is expensive to ship, hard to demo, or still on the roadmap, this is the case for turning your CAD data into a sales asset. Read the full whitepaper

Industrial Trade Shows: 3D Visualization Is Changing the Game
Immersive
August 20, 2026
Industrial Trade Shows: 3D Visualization Is Changing the Game

Next-generation Extended Reality tech is raising the standard for industrial 3D visualization. Complex equipment becomes much easier to show customers and walk them through, without hauling a physical unit to every meeting. For industrial machinery manufacturers, that’s real business value: by reducing their dependency on physical equipment, companies can improve sales demonstrations and make customer education more visual, scalable, and effective. The value is clear at trade fairs. Industrial machinery manufacturers can usually show only one machine in one configuration. On top of that, the most important parts are often hidden inside, and the product is displayed without the environment it was designed for. So the exhibitor arrives at the show unable to demonstrate the product they’re selling. This is where a new generation of Virtual Showrooms comes in. We’ll look at how it solves that problem below. The problems manufacturers face at trade shows But first, let’s look at the main challenges manufacturers face when relying on traditional trade show methods. The product does not fit. What a manufacturer sells is often far too big for any booth. A single production line can fill half a plant, and one machine can weigh five tons. But the trade show offers thirty square meters of rented floor at a high price. Exhibit space is the single largest line in the entire exhibiting budget: for U.S. B2B shows it accounts for 40.5% of an exhibitor’s total spend, more than any other category. At Hannover Messe (Germany), a modest 18 to 36 square meter stand costs €18,000 to €45,000, and a large one, 72 to 200 square meters, can pass €250,000. The equipment was never built to be lifted onto a stand, so a company pays the highest price in its budget for the one thing that still cannot hold the product it came to sell. Shipping machinery is expensive. The heavier the machinery, the higher the sales costs. Transportation, on-site assembling, and travel expenses can run into six figures per event. On average, construction, logistics and assembly add up to around €100,000 to get a single product onto the floor, and the largest, most impressive machines, the ones a company most wants buyers to see, are precisely the ones that push that figure highest. Buyers see a sample, not the range. Manufacturers have to choose which models to bring, and buyers never see the full portfolio. They are also limited by demo capabilities: system flow, scale, and engineering remain hidden. The value is hidden under the housing. Much of what a manufacturer charges for happens inside the machine, out of sight. Visitors stand in front of a closed housing and see only the outside of a product whose real engineering is sealed away. There is no context. Buyers need to understand how it fits into their own production line, works with existing equipment, and supports their process. That context cannot be brought to the booth, so much of the product’s value has to be imagined. The product may not exist yet. Sometimes the product a company needs to sell is still on the roadmap, or only half-built in a workshop. Industrial sales cycles are long, and buyers often need to commit before a machine physically exists. You are competing for the same buyer as everyone else. A trade show gathers the entire industry into one hall, which means every rival is only a few steps away. At IMTS 2024, visitors had to work through 1,737 exhibitors spread across more than 1.2 million square feet. Hannover Messe 2025 was larger, with 3,694 exhibitors from 62 countries and more than 123,000 visitors. In a room like that, you have to stand out to get noticed. How 3D solves each problem So, how can 3D visualization help? Problem: The product is too large to ship, or it does not exist yet. With XR, prospects can walk around a full-scale model of a product, explore different options for an engineered-to-order build, and see exactly what they are buying months before the first unit is produced. This can shorten the sales cycle. The format changes, but the product remains in front of the customer. Problem: Shipping and floor space eat the largest share of the budget. Extended Reality replaces the heaviest line items in the industrial sales budget with a single portable asset. Each of the costs for equipment transport, exhibition stand rental, assembly crew, technical team travel recurs with every event. Over a year, they compound into one of the largest budget lines in industrial sales. A 3D model is created once and can be presented on a screen, in AR, or in VR using equipment that occupies only a small part of the stand. It can be updated when the product line changes and redeployed without additional logistics. Problem: A physical demo can show only one configuration. Let each visitor build their own version. For configurable machines built to order, each customer can preview their exact configuration in detail rather than a generic prototype. A configurator on a touchscreen, or the same system in AR or VR, lets them choose components and options, see the result update instantly, and leave with a build that matches their plant. One asset covers all two hundred configurations in the catalog. Problem: The most important parts of the product are hidden inside. Exploded views, cutaways and animations show what happens beneath the surface. On a screen, a visitor grasps how the system works within seconds; in AR or VR, they can pull it apart and study the internals in detail.  Problem: Buyers can’t easily picture how the product fits their own plant. VR can put a buyer inside a virtual production line. AR can drop the equipment at full scale into the room right in front of them, so they can walk around it. This way, they can see how it fits their facility instead of guessing. Problem: You have a few minutes to catch attention, and competitors are right next to you. Big screens with motion are…

The State of 3D Medical Image Visualization in 2026
Immersive
June 29, 2026
The State of 3D Medical Image Visualization in 2026

Today’s imaging systems are more powerful than ever. A single CT scan generates hundreds of cross-sections. An MRI cardiac study captures the heart in four dimensions. A full-body PET produces a dense volumetric map of metabolic activity across every organ system. And yet, in most hospitals today, clinicians consume all of that data the same way they did in the 1990s: as 2D slices, scrolled one frame at a time, with the third dimension reconstructed entirely in the radiologist’s head. That gap between the data that exists and the data that gets used is what 3D medical visualization is closing. Progress hasn’t been uniform. The specialties with the highest spatial stakes have moved fastest. In oncology, where tumour margins and vascular relationships determine whether a resection is safe, 3D visualization is now routine. In cardiology, where structural defects live in three dimensions that 2D echo can only approximate, volumetric review has become standard practice for complex case planning. For these teams, rotating a segmented model or flying through a volume-rendered vessel is part of the reading workflow. Within healthcare, oncology drives roughly 34% of total 3D imaging spend: 52% of cancer centers already use 3D imaging as part of their standard workflow, and 44% of cardiology departments do the same. For much of medicine, the shift is still underway. But the direction is clear. The market reflects it. The global 3D medical imaging market was valued at $21.43B in 2025 and $23.39B in 2026 and is projected to reach $42.75B by 2032  at a compound annual growth rate of 10.36%. Healthcare has become the largest adopter of 3D imaging technology overall. In this article, we break down what 3D medical visualization actually means technically and where it creates measurable clinical value. The imaging data problem Begin with the scanners, because they don’t produce data the same way: CT measures X-ray absorption, so dense tissue like bone reads strongly while soft tissue stays faint: the default for trauma, lung, and skeletal work. MRI reads tissue magnetic properties instead of density, trading speed and bone detail for soft-tissue contrast nothing else matches. PET maps metabolic activity rather than structure, and almost always travels fused to a CT or MRI so the active regions have anatomy to sit against. Ultrasound produces a live volume but depends heavily on probe angle and operator skill. Cone-beam CT gives a tight, high-resolution field at the cost of coverage, which is why it dominates dental and interventional suites. All of these imaging methods capture a 3D volume of the body. Yet in most cases, doctors still review that data as a series of 2D slices. At first glance, this seems surprising: why collect rich 3D data only to view it in 2D? Part of the answer is habit and established workflows, but there are also practical reasons why 2D slices remain the standard in medical imaging. Raw data, nothing interpreted. A slice shows the scan as acquired. Every 3D rendering is the product of decisions which densities to display, which to hide, where to set the threshold and any of those can suppress a real finding or manufacture one that isn’t there. Full coverage of the dataset. Scrolling slices walks the eye across every voxel in the study. A 3D view by definition hides whatever sits behind the surface it shows, and for catching a small lesion or a faint ground-glass opacity, seeing everything matters. 3D earns its place once the task moves past detection: Spatial relationships. 3D visualization makes it easier to understand how anatomical structures relate to one another. Instead of mentally reconstructing anatomy from dozens of 2D slices, clinicians can view organs, vessels, and abnormalities as a single 3D model. Change over time. Tracking changes across multiple scans becomes much easier in 3D. By measuring the volume of a structure over time, clinicians can quickly identify trends that may be difficult to spot in individual slices. Communication. A 3D model is something a patient, a referring physician, or a multidisciplinary team can read at a glance, where a slice stack means little to anyone outside radiology. So, 3D visualization is most valuable when understanding spatial relationships is difficult or time-consuming in 2D. What complicates this in practice is the format the data arrives in. Most medical imaging is still stored as DICOM, a standard built around 2D-image workflows. DICOM is the backbone of medical imaging, but several of its legacy choices make 3D visualization and analysis harder to build on top of it. Gathering everything a full analysis needs is one problem: a careful read of a pathology usually draws on prior scans and the patient’s imaging history, and that data sits scattered across separate studies and series rather than in one place. Interoperability is another. DICOM has to exchange data with the hospital’s other systems, such as PACS, RIS, and the electronic health record, and every connection point adds friction. The input itself is uneven too: scans vary in quality and completeness depending on how and where they were acquired, so a tool built for real cases has to hold up across that range. We’ve written separately about why DICOM is stuck in the ’90s. What “3D medical visualization” actually means There are five techniques in common use. Most clinical software uses two or three of them together. Segmentation comes first, because the others depend on it. Segmentation. Something has to label what is in the scan before the rest can work. It needs to know which voxels are liver, which are tumour, which are vessel wall. This used to be manual work. A radiologist drew outlines on each slice, which for a complex case could take close to an hour. Two radiologists rarely produced identical outlines. AI tools changed this. TotalSegmentator and similar models label most organs in a CT scan in under a minute. The clinician checks and corrects the result instead of drawing it. This is what makes the other four techniques practical for routine use. Multiplanar reformatting (MPR)….



Let's discuss your ideas

Contact us