The 4 Types of Augmented Reality and What Each Is Good For

Augmented reality gets talked about as a single technology. It is four different things with different costs, different hardware needs, and different reasons to fail.

Knowing which type you need is most of the work. Here they are.

1. Marker-Based AR

The camera recognizes a specific image or code and places content relative to it.

Point a phone at a printed marker and a 3D model appears anchored to it. This is the oldest and most reliable type, because the anchor is unambiguous.

Good for: packaging, print advertising, museum exhibits, trade show materials, instruction manuals.

Cost: lowest of the four. The tracking problem is solved and the tooling is mature.

Limitation: requires the physical marker. No marker, no experience.

2. Markerless AR

The device understands the surfaces in the space and places content on them without a predefined anchor.

This is what furniture placement apps use. The phone identifies your floor and puts a sofa on it.

Good for: product placement and try-on, the highest value commercial applications available today.

Cost: moderate. Requires accurate 3D assets, which is usually the real expense.

Limitation: tracking quality varies with lighting and surface texture. Performance in a dim room is noticeably worse.

3. Projection-Based AR

Light projected directly onto physical surfaces, with no device held or worn.

Assembly guidance projected onto a workbench. Interactive floor displays in retail.

Good for: industrial assembly, guided manufacturing, permanent retail installations.

Cost: high, and it is fixed installation cost rather than software cost.

Limitation: location specific and sensitive to ambient light. It does not travel.

4. Superimposition AR

Replacing or augmenting a view of an object with additional information. Medical imaging overlaid on a patient, a schematic overlaid on a machine.

Good for: maintenance, medical, complex repair work.

Cost: highest. Requires accurate recognition of the specific object and reliable registration.

Limitation: precision requirements are severe. Being slightly off is worse than not doing it.

How to Pick

Start from the constraint rather than the technology.

If the customer will encounter a physical printed thing, marker-based is cheapest and most reliable.

If they need to see how something looks in their own space, markerless. This is the one with proven commercial return, and if you sell physical goods it is where I would start.

If the work happens at a fixed station, projection.

If a trained person needs information registered precisely onto equipment, superimposition, and budget accordingly.

Why AR Projects Fail

The most common failure I see is picking a type for its impressiveness rather than its fit. Superimposition demos beautifully and fails in production because registration accuracy in the real world is hard.

The second is underestimating the 3D asset cost. Every type needs accurate models, and companies consistently budget for software and not for the assets.

The third is requiring an app download. A web-based AR experience that opens from a link converts vastly better than one requiring an install, and WebXR support has gotten good enough that the app is usually unnecessary.

Where This Is Heading

The practical near-term value stays in markerless placement on phones, because that is the device everyone has. I made the broader argument about that in spatial computing and marketing.

The low regret investment is accurate 3D models of your products. They are useful today and they are the input to whatever comes next.

For the technical standards, the WebXR Device API documentation is where the browser-based side is defined, and it is the part worth tracking.

Do You Need an App to Use Augmented Reality?

No, and requiring one is the most common reason AR projects underperform. Web-based AR opens from a link in a mobile browser, and it converts dramatically better than anything requiring an install, because you lose most of your audience at the app store.

WebXR support in mobile browsers has become good enough for the commercially useful cases: placing a product in a room, viewing a model at scale, simple try-on. That covers the applications with proven return.

When an app is genuinely warranted: when you need sustained use over time, access to device capabilities the browser restricts, or offline operation on a factory floor. Those are real cases and they are narrow.

The practical implication for a marketing project is a link you can put in an email, a QR code on packaging, or a button on a product page. That distribution flexibility is worth more than the marginal quality an app would add.

One thing to test before committing: performance on mid-range phones rather than the newest device in your office. AR is demanding, and an experience that runs smoothly on a current flagship can be unusable on the three-year-old phone a large share of your customers carry. That gap has killed more AR projects than any technical limitation of the format.

The most common failure across all four types is the same: picking the one that demos best rather than the one that fits the constraint. Superimposition is the most impressive in a controlled demo and the hardest to make reliable in the field. Markerless placement is unglamorous and it is where the measurable commercial returns actually are.

The second most common is underbudgeting the 3D assets. Every type needs accurate models, companies consistently budget for software and not for the models, and the project stalls at the point where somebody has to produce them.

The low regret investment underneath all four types is the same: accurate, optimized 3D models of your products. They are useful today for phone-based placement and they are the input to whatever device comes next, which is the argument I made in spatial computing and marketing.

Which type does your use case actually need? Most of the projects I have seen picked the wrong one before asking.