4 Ways to Improve CAD-AR Overlay Accuracy | Considerations for On-site Implementation

Table of Contents

• Why accuracy matters in CAD-AR overlay

• Prerequisites that affect the accuracy of CAD-AR overlay

• Method 1: Unify reference points and reference lines

• Method 2: Organize data to include only what is necessary for on-site verification

• Method 3: Fix height, orientation, and inspection viewpoints in advance

• Method 4: Manage the entire workflow from on-site verification to reflecting corrections

• Field implementation note 1: Consider required accuracy by use case

• Field implementation note 2: Question discrepancies between as-built conditions and drawings

• Field implementation note 3: Eliminate differences in how stakeholders view the overlay

• Approach to embed CAD-AR overlay into on-site practice

• How to further improve accuracy and connect it to practical work

Why Accuracy Is Important in CAD AR Overlay

CAD AR overlay allows drawings or model information to be superimposed on the actual site scenery or existing structures, making it a useful method for pre-construction review, stakeholder briefings, comparisons with existing conditions, and verification of changes. A major advantage is that it shortens the cycle of looking at drawings on a desk, going to the site, and returning to the drawings for interpretation, and it makes it easier to align spatial understanding. However, when used in practice, merely being able to display the overlay is not enough. The overlaid position must feel convincing, multiple viewers must be able to interpret it the same way, and the accuracy must be sufficient to support the necessary decisions.

On site, even if something works on the drawings, the impression can change when you actually stand in the space. It is not uncommon for the distance to existing structures to be too small, for there to be more of a sense of confinement than expected, for aisles or work flows to be affected, or for inspections and future maintenance to look difficult. The value of CAD-AR overlay is that it can detect these kinds of discomforts at an early stage, but if the overlay’s accuracy is low, it can instead cause hesitation in on-site decision making. Even a slight misalignment can make it unclear how much to trust it, and in the end you revert to checking the drawings.

Also, the reason accuracy becomes important is that there isn’t just one purpose for viewing things on site. Pre-construction position checks, clash detection, visualization for explanations, comparisons of changes, sharing as-built imagery, and so on — the level of precision required varies depending on the use. What holds true across uses is that if the alignment is unstable, both explanations and verifications take more time. On site, the more a system looks convenient, the more likely it is to fall out of use if its behavior is even slightly unstable. That is why reproducible accuracy on site is far more important than flashy visuals.

Furthermore, CAD/AR overlays are also a means of aligning stakeholders' understanding. Designers place importance on the consistency of drawings, contractors emphasize constructability, and managers may prioritize maintenance and safety. When discussions are held based on drawings alone, even when everyone is looking at the same subject the focus tends to diverge. Being able to view design information overlaid on the site makes it easier to align the assumptions underpinning discussions, but for that to work, at minimum the way reference points are defined and the information that is displayed must be stable. Accuracy is not simply about reducing numerical errors; it is also about creating a condition that can be trusted and used on site.

Many practitioners who search for "CAD AR overlay" want to know not only how to display it but also how to reduce positional misalignment and the sense that something looks off. In fact, the parts that tend to cause trouble in on-site implementation are upstream of the operation itself. If operational design—such as how to establish reference points, what to show, where to verify, and how to reflect corrections—is weak, then no matter how good the appearance, it becomes difficult to use in actual work. Improving accuracy should be considered not as a problem solved by display technology alone but as arranging verification conditions that can be reproduced on site.

Prerequisites That Affect the Accuracy of CAD–AR Overlays

Before considering methods to improve accuracy, the premise you should first grasp is where the accuracy of CAD AR overlays is determined. On site, when something appears misaligned people tend to assume it's a problem with the device or the display functionality, but in reality the cause is often at an earlier stage. Insufficient organization of drawing data, ambiguity in reference points, inconsistencies with current site conditions, differences in how elevation information is handled, and differing perspectives among stakeholders can combine and manifest as positional misalignment on site. In other words, the problem is that the conditions for overlaying are not in place prior to the display itself.

What's particularly important are the differences in reference points between the drawings and the site. Even if the drawing uses a certain intersection or origin as the reference, on site people may rely on a different corner or another landmark. As a result, although the overlay may appear to align at the moment it's placed, moving slightly can create a sense of mismatch, or someone else may perceive it as misaligned. The accuracy of AR overlay depends heavily on where you choose the starting point and which line you use as the alignment reference. If this is left ambiguous, the appearance will vary each time and on-site reliability won't improve.

Being satisfied with alignment only in plan view can also cause a loss of accuracy. On site, not only left-right and front-back positions, but also height, orientation, tilt, and angular relationships with existing elements are important. When overlaying based on plan drawings, the overall position may seem to match, but heights can differ, items can appear slightly rotated, or the sense of depth can feel off. If these are used for construction or clash checking, such discrepancies can lead to major rework in later stages, so they need to be addressed and resolved early on.

Furthermore, how you think about accuracy changes depending on the intended use. When used for on-site explanations or sharing the final image, ensuring a common spatial understanding is important, and clarity may be prioritized over minor inaccuracies. On the other hand, when used for position checks or interference checks, the strictness of reference alignment becomes more important. Trying to make a solution universally applicable for every use can make preparations overly burdensome and, in the end, it will not be used on site. To improve accuracy, it is essential to first clarify the verification conditions required for each use and then design operational procedures that meet those conditions.

Thus, the accuracy of CAD-AR overlays is not determined by a single technical element. It is determined by how drawing data is prepared, how on-site reference points are established, how height and orientation are handled, and how stakeholders’ understanding is aligned. If you want to improve accuracy in field implementation, you need to review the foundation—what assumptions you use when overlaying—rather than focusing only on comparing display functions. From here, we will examine in detail four methods for establishing those assumptions in practice.

Method 1: Unify the reference point and reference line

One way to increase accuracy is to standardize the reference points and reference lines. What matters most in CAD–AR overlays is where you link the drawing to the actual site. If this is ambiguous, even when using the same data each operator will interpret it slightly differently, causing positional discrepancies. Conversely, if the reference points and reference lines are clearly defined, on-site reproducibility improves and the burden of re-aligning during each check is reduced. More than visual clarity, agreeing on which point to use as the origin is the starting point for improving accuracy.

A suitable reference point is a location that is easy to find on site, unlikely to change, and readily recognized in the same way by multiple people. For example, a building corner, the end of an existing structure, a floor reference line, or the intersection of a clearly defined centerline are candidates. Even if a point is convenient on the drawings, it becomes a weak reference if it is hard to see on site, hidden by temporary materials, or difficult for different people to identify as the same spot. If you want to stabilize alignment accuracy, you should choose reference points that prioritize on-site reproducibility rather than convenience on drawings.

Also, it is safer not to rely on a single reference point. If you align using only one location, any misidentification of that point or difference from the actual conditions will become a global error. If you prepare multiple reference points or reference lines, it becomes easier to reconcile orientation and rotation, and any sense of mismatch when you move slightly is reduced. Especially on-site, because inspections are not done only from a fixed frontal view but also by walking around or by multiple people viewing from different positions, checking alignment against multiple references is extremely effective in practice.

Even more important is not to leave the concept of reference points and reference lines solely in the minds of individual staff members. Even if the designer understands them, if field personnel or other stakeholders are looking at different references, discrepancies in understanding will arise on site. You need to share in advance which points will serve as references and which lines will be aligned so that anyone checking can align them the same way. This is not a matter of operational procedures, but of standardizing the verification conditions. If AR overlays feel unstable on site, it is not uncommon for a lack of shared reference standards to be the underlying cause.

Unifying reference points and reference lines is not a flashy improvement, but its effect is very significant. Whether for on-site explanations, comparative checks, change verification, or sharing as-built imagery, the visual stability depends on whether this foundation is in place. Before looking for new features to improve accuracy, first standardizing the references is the most cost-effective improvement.

Method 2 Organize only the data necessary for on-site verification

The second method is to pare the data down to only what is necessary for on-site verification. CAD drawings and models contain a lot of information, and that is necessary to maintain the accuracy of drafting and design. However, when used for AR overlays on site, displaying all of that information as-is can obscure the elements you want to check, and as a result both alignment and judgment become unstable. To improve accuracy, it is important to reduce the displayed information and make the reference and the target being checked clearly visible.

For example, if the objective is to verify plan position, prioritize centerlines, primary outlines, reference lines, and the positional information of the elements to be checked. If the objective is to perform an interference (clash) check, make elements related to surrounding equipment, work movement paths, and clearances easy to see. If the goal is to share a finished image, retain information related to shape and appearance and suppress other details to make it easier to understand. Because the information required differs by purpose, using data organized to match on-site verification makes it easier to balance overlay accuracy and clarity.

Also, organizing data helps you discover positional misalignments more quickly. When there are many unnecessary lines and annotations, it becomes hard to tell which elements should serve as references, and each person ends up looking at different places. As a result, even if they think they are looking at the same screen, their judgments about what is aligned and what is off will differ. On site, time is limited, so being able to find important information immediately directly affects the accuracy of verification.

Furthermore, the data used on site may differ from the layout that is easy to view in the office. While it may be more convenient in the office to stack drawing layers to grasp the whole picture, showing too much information at once on site can actually make judgment more difficult. Because AR overlays place information into the physical space, they can cause the eye to wander more than drawings do. For that reason, it is important to keep the site in a state where it is intuitively clear what should be looked at right now.

Organizing data down to only what is necessary for on-site verification does not mean sacrificing accuracy. Rather, by narrowing what is shown, you improve decision accuracy on site. The value of CAD-AR overlays is not in the amount of information they can display, but in converting that information into a form that is easy to use on site. If you want to prevent positional misalignment, start by reducing the information displayed and clarifying the reference and the item to be checked.

Method 3: Fix the height, orientation, and inspection viewpoint first

The third method is to fix height, orientation, and the verification viewpoint in advance. When using AR overlays on site, it's easy to focus only on aligning planar positions, but in reality offsets in height and orientation and differences in viewing position can greatly influence perceived accuracy. Even if something appears to be aligned in plan, moving slightly to the side or changing your eye height can suddenly create a sense of mismatch. This tends to happen not so much because of a problem with the overlay itself, but because the conditions under which it is being checked are not fixed.

Particular aspects to pay attention to in height information are the assumptions about floor level, finished surface, the heights of foundations and support frames, and the mounting heights of existing equipment. Even if these are shown as numbers on drawings, the reference surface may not be shared on site. If you overlay drawings without that shared reference, something that seems fine in plan view can turn out to be too high, too low, or have a different height relationship with existing equipment in the field. Especially for equipment replacement or renovation, even a slight difference in elevation can create a significant sense of incongruity, so height should be regarded as important as the plan layout.

How you set orientation is equally important. If you don't decide in advance which direction to use as the reference—north on the drawings, the building's axis, the street centerline, the front of equipment, etc.—slight differences in rotation will manifest as an overall misalignment. This kind of misalignment is hard to notice when looking at a single point, but becomes very noticeable from a distance or from another angle. Aligning orientation is not just a matter of simple rotational correction; it also relates to the idea of what is treated as the "front" on site.

Also, fixing the inspection viewpoints in advance is highly effective for on-site implementation. Although it might seem ideal for something to be understood the same way from any vantage point, in practice it is more efficient to decide beforehand which viewpoints have the highest verification value. For example, deciding viewpoint scenarios according to use—how it appears from an aisle, from the work position, or from the user's perspective—makes checks less likely to vary. Rather than showing things while moving around freely, narrowing down the positions to be viewed in advance improves both the accuracy and speed of verification.

Fixing height, orientation, and the verification viewpoint in advance may feel restrictive, as if imposing conditions on the overlay. However, in practice this organization is necessary to stabilize how things appear on-site and to make explanations reproducible. If you want CAD/AR overlays to be used for on-site decision-making rather than ending as mere visual novelty, being able to review them under the same conditions is more important than allowing free viewing. Preventing positional drift is not just about perfectly tracking while moving; it is also about creating a state in which you can stably verify from critical viewpoints.

Method 4 Manage the entire process from on-site verification to implementing corrections

The fourth approach is to manage the entire process from on-site verification through implementing corrections. AR overlays may seem to create value at the moment they are viewed on site, but in practice how those results are handled afterward is extremely important. Even if overlaying on site reveals insights—such as a feeling that a position is off, a constrained relationship with existing installations, or problems with how reference points were established—if those insights end there on the spot, the same issues will be repeated at the next check. To improve accuracy, it is necessary to create a workflow that feeds the results of on-site verification back into data improvements.

First, the important thing is to document which reference you used to check and where you felt something was off. Simply saying that the appearance was slightly misaligned will not lead to further improvements. If you organize at which position, from which viewpoint, and against which standard you noticed what kind of discrepancy, it becomes easier to distinguish whether the issue lies with the drawings, with a difference from the actual conditions, or with how the reference was taken. This is necessary to prevent accuracy from ending as a matter of subjective feeling and to turn it into information that can be used for improvement.

Next, a process is needed to incorporate the insights gained on site into the official drawings and models. On site, conditions that were not visible at the design stage may be discovered. If issues such as the existing installation being slightly off, the way things appear changing under temporary conditions, or unexpected interferences are shared only on the spot and not fed back into the updated data, the same problems will arise next time. The more the workflow connects on-site verification with the reflection of corrections, the more stable the accuracy of AR overlay becomes with each iteration.

Also, when the workflow for applying corrections is well established, it becomes easier to compare and verify the before-and-after changes. Because you can trace which parts changed, why they changed, and which verification conditions are affected, on-site explanations can be given more quickly. Conversely, if the update history remains ambiguous, it becomes unclear which information is official, and the reliability of AR displays declines. In on-site implementation, the freshness of the display and the clarity of the differences matter more than you might think.

Managing the entire process from on-site verification through reflecting corrections may seem to increase the amount of preparation required. In reality, however, it reduces the waste of repeating the same checks multiple times and minimizes differences in understanding among stakeholders. If you truly want CAD AR overlays to take root on site, you must not make it a process that ends with simply viewing them there. Only when what is observed is turned into operational steps that lead to improved accuracy does the system become meaningful in practice.

On-site implementation note 1: Consider the required accuracy separately for each use case

From here, we will look not only at the methods but also at precautions for on-site implementation. The first point to consider is to separate required accuracy by use case. Because AR overlays can seem applicable to everything, it is common when introducing them to try to cover everything with a single workflow. However, the way required accuracy should be considered differs for on-site explanations, sharing the finished-image, pre-construction clash checks, and position verification. If you proceed without distinguishing these, you can end up with heavy preparation and a system that is difficult to use on site.

For example, if you only want to share how it will look when completed or to convey spatial awareness, grasping the overall positional relationships and sense of volume takes priority. On the other hand, if you want to verify clearances from existing elements or the validity of installation positions, the precision of reference points and height/elevation conditions becomes more important. It is not uncommon for a visualization that was sufficient for explanatory purposes to be inadequate for positional verification. By separating the required accuracy by use case, you can avoid forcing all requirements into a single overlay.

Also, by separating uses, it becomes easier to narrow down the data that needs to be prepared. Since the inspection targets, reference points, viewpoints, and necessary comparison items become clear, it also becomes easier to understand what to look for on-site. This not only improves accuracy but also raises the quality of on-site explanations and meetings. In practice, an approach that ensures the accuracy required for the purpose is easier to sustain than trying to make everything highly precise.

Furthermore, by separating operations according to use case, on-site personnel can more easily understand when to use it. If it’s clear whether it’s being used for explanation, for interference checking, or for comparing changes, the field can use it without hesitation. In sites where AR use does not become established, this ambiguity about when to use it is often an obstacle.

Thinking about required accuracy separately for each use is not a compromise of accuracy. Rather, it is an approach to provide the precision the site needs—no more, no less. To avoid failures in on-site implementation, it is important not to aim for a one-size-fits-all usage, but to separate the verification conditions required for each use and establish reliable operations for each.

On-site implementation precaution 2: Question discrepancies between the actual conditions and the drawings

The second point to watch for is to question differences between the as-built conditions and the drawings. When an AR overlay doesn't align properly, it's easy to assume the problem lies with the display or the alignment method, but in reality the drawings and the actual site may already be out of sync. Especially on sites with many renovations, upgrades, or existing installations, it's not uncommon for the site to no longer match the drawings. Additions of temporary materials, updates to existing equipment, finish changes, or subtle shifts can all create a noticeable discrepancy when overlaid. If you attribute all of that discrepancy to the AR side, you're likely to pursue the wrong direction for improvements.

Questioning discrepancies with the actual site conditions is a critically important starting point for on-site verification. Drawings are, at best, information from the design or record stage, and the site can continue to change afterward. Especially for equipment that has been in use for a long time or locations that have undergone multiple renovations, drawings may not be a complete representation of the current conditions. To trust and use AR overlays, you must first take the stance of verifying how accurate the drawings are relative to the current site conditions.

Also, by questioning differences from the current/as-built condition, it becomes easier to organize the sense of mismatch when overlaying. Look at where things appear to be offset, whether that offset is in one direction, whether it occurs only with specific existing elements or across the whole model — doing so makes the cause easier to see. If you can separate whether it’s a reference point problem, an omission in updating drawings, or a change in site conditions, corrective actions can be taken more quickly. In field implementation, the ability to make this distinction determines how easy continued use will be.

Furthermore, by assuming differences from the current state, the way AR is used becomes more practical. Rather than expecting a perfect match from the outset, it becomes easier to use it first as a tool to find discrepancies and then connect that to the necessary corrections and verifications. With this approach, on-site personnel are less likely to view AR as a cure-all and more likely to adopt it positively as a practical aid.

Questioning discrepancies between the as-built conditions and the drawings does not undermine the reliability of AR use. Rather, finding discrepancies early brings the overlay closer to being robust enough for practical use. What truly matters in on-site implementation is not assuming deviations are zero, but having a clear understanding of what to suspect and how to correct them when deviations occur.

On-site Implementation Consideration 3 Eliminating Differences in Stakeholders' Perspectives

The third point to note is to eliminate differences in how each stakeholder perceives things. AR overlays are used to show information on site and are often thought to be easier to understand than drawings alone; however, in practice, if who is looking at what is not aligned, understanding will diverge. Designers often focus on dimensional correctness, contractors on workability, and managers on maintainability. Even when looking at the same screen, if they pay attention to different elements, their conclusions will differ. When implementing on site, operations need to be organized with these differences in mind.

To do that, it is important to clarify what kind of review the session is for before overlaying. If it is clear whether the purpose is position checking, clash checking, explanation, or change comparison, stakeholders can more easily align on the points to look at. Conversely, if you bring something to the site in a state where anything can be viewed, discussions tend to branch into different topics depending on each party’s perspective, and confirmations become scattered. In sites where AR feels unhelpful, this weak setting of objectives can be the cause.

It is also effective to align in advance the order of on-site explanations and the viewpoints to be used. If it is shared where to look from, which criteria to reference, and what to observe to consider verification complete, stakeholders will find it easier to discuss the same issues. The advantage of AR is that it enables discussion over a shared view, but to leverage that advantage the verification/acceptance criteria must also be shared.

Furthermore, eliminating differences in how each stakeholder perceives things also improves the accuracy of judging positional misalignment. When one person thinks it’s correct and another thinks it’s off, on-site decision-making comes to a halt. If the evaluation criteria and the acceptable degree of alignment are shared, discussions about discrepancies become more concrete. This also has a significant effect in speeding up on-site verification.

Eliminating differences in how stakeholders perceive things is not a matter of technology but of on-site operations. However, this point is crucial for embedding AR into everyday practice. Improving overlay accuracy is not just about tightening the numbers; it also means creating a state in which everyone can use it with the same understanding. If you want AR to be truly usable in the field, you should design it to include shared visual understanding.

How to Establish CAD-AR Overlays On-site

Making CAD-AR overlays stick on site requires more than just adopting the technology. If it isn’t clarified within the site workflow when, who, and for what purpose will use it, it may be a hot topic at first but difficult to sustain. What sites where it has taken hold have in common is that AR is not a special way of showing things but a natural part of verification tasks. To create that state, it’s important not to try to change everything from the start, but to begin with situations where the effects are easy to see.

For example, it's easier to introduce by starting with use cases where the value is clear: situations where confirming positions relative to existing installations takes time each time, situations where it's difficult to get everyone aligned with drawings alone when explaining to stakeholders, and situations where checking the impact of changes is laborious. Trying to use it for all drawings, all models, and all processes from the outset increases the preparation burden and makes the site more likely to see it as a hassle rather than a benefit. You're more likely to succeed if you start small and find approaches that are meaningful on-site.

It is also important to treat usability complaints from the field as material for improvement. Remarks such as "hard to see," "standards are unclear," "too much drawing information," "difficult to compare," and "hard to track corrections" are not complaints about the technology but hints for translating it into practical use. If these are ignored, AR will end up merely as a means of display. An operational practice that takes root will be gradually refined by collecting and addressing the field’s sense of discomfort.

Furthermore, it is important to make the benefits for each stakeholder visible. Construction management gains the advantage of reduced verification time; on-site personnel gain the advantage of easier spatial awareness; and designers gain the advantage of being able to more quickly grasp discrepancies with actual site conditions. When these concrete benefits are shared, on-site acceptance is more likely to advance. A system that is convenient for only a subset of people will not last long on site.

When considering adoption, the most important thing is not to make the introduction of AR an end in itself. If the true objectives—reducing on-site waste, reducing discrepancies in understanding, and speeding up verification—are clear, the necessary preparations and areas for improvement become easier to identify. To establish CAD-AR overlays on-site, it is important to position them as tools to reduce on-site concerns and rework, rather than focusing on the appeal of the technology.

How to further improve accuracy and apply it in practice

As noted above, improving the accuracy of CAD-AR overlays requires standardizing reference points and reference lines, organizing data for field use, fixing heights, orientations, and verification viewpoints, and managing the entire process from on-site verification to applying corrections. Furthermore, in field implementation it is essential to separate required accuracy by use case, be skeptical of differences between the model and the existing conditions, and align how stakeholders perceive the overlay. By putting these elements in place, AR overlays become not just display tweaks but practical tools that effectively support on-site verification.

What you should pay particular attention to is the perspective of reconciling improved accuracy with ease of use in the field. If you try to be exact about everything, preparation becomes burdensome and it stops being used on site. Conversely, if you prioritize ease of use alone, you can end up with something that, although visible, cannot be used for decision-making. What has value in practice is the ability to deliver the necessary accuracy where it is needed and to be usable repeatedly in the field. For that reason, the stability of standards and procedures is more important than flashy displays.

Also, to truly connect AR overlays to the field, it’s important not to treat visual inspection and position verification as entirely separate. When something feels off on site, being able to turn that feeling into a more reliable position check makes it easier to move beyond mere explanation and reach construction decisions. In particular, when you need to improve the reproducibility of positions on site, it’s effective to pair AR-based spatial understanding with an approach for handling positioning.

When considering such operations, options include means to incorporate high-precision positioning in a form that’s easy to handle on site, such as LRTK (an iPhone-mounted GNSS high-precision positioning device). If AR overlays speed up on-site understanding while allowing, as needed, transition to high-precision position verification, the value of field implementation is likely to increase further. If you want CAD AR overlays to be more than a technology that only makes things visible and instead a system that can be used on site without hesitation, it is important to arrange the workflow so that viewing drawings, overlaying and understanding them on site, and confirming the necessary positions are integrated into a single flow.