A construction drawing is never only a drawing. A single sheet can contain a dimension that alters a structural connection, a note that changes a fire-rating requirement, a revision cloud that affects procurement, or a layer hidden beneath dozens of others. Multiply that complexity across architectural, structural, MEP, civil, and specialist drawings, then add specifications, contracts, requests for proposals, and BIM files. The result is a document universe that design teams must inspect under constant time pressure.
South Korean startup Searchdog says it has built a system to make that universe searchable. The company claims its platform can reduce the time contractors and design firms spend reviewing whether blueprints comply with client specifications and building codes by up to 70%. The platform analyses design, procurement and construction documents, including 2D and 3D drawings and BIM information.
Searchdog’s claim is ambitious. Yet the more interesting question is not whether a machine can “read” a drawing in the human sense. It is whether it can help project teams find the small, expensive details that get buried in the information overload of modern construction.
A drawing is a database in disguise
Architects read drawings spatially. Engineers read them technically. Contractors read them operationally. Quantity surveyors read them commercially. Regulators read them for compliance. Each professional sees a different layer of risk in the same plan.
A typical set of documents includes floor plans, sections, elevations, schedules, construction details, finish plans, reflected ceiling plans, coordination drawings, specifications, addenda and revision records. These files contain written instructions, geometric relationships, symbols, dimensions, line types and references to other documents.
That complexity creates a serious review burden.
Searchdog began by handling contracts and requests for proposals before extending its system to two-dimensional and three-dimensional blueprints and BIM. The company says it extracts geometric, layer, and dimension data from documents and drawings, then structures and links this information through search technology.
This distinction is important. A conventional document-search tool may identify words such as “fire door,” “concrete” or “Level 03.” A design-review system needs to understand where those words appear, which objects they refer to, which drawing version is current and whether the stated requirement matches the geometry or data attached to the model.
In other words, it must turn drawings into a connected dataset.
The hidden cost of file conversion
Construction projects use an uncomfortable mix of file formats. Design teams may work across proprietary CAD, BIM, PDF, spreadsheet, point-cloud and document systems. Autodesk, Bentley Systems and AVEVA each operate within different software ecosystems and blueprint formats.
Searchdog founder and CEO Baek Jun-sun said that converting these files into a common format can result in the loss of about half of the original information. That is more than an interoperability inconvenience.
When geometry, layers, dimensions or object relationships disappear during conversion, the model loses context. A wall may become a collection of lines. A tagged object may lose its associated data. A revision layer may vanish. A construction note may detach from the detail it was meant to clarify.
For computational design and BIM professionals, this is a familiar frustration. Information-rich models are valuable only when their information survives movement between platforms.
Searchdog’s stated approach is to preserve and structure critical drawing information rather than flatten it into a generic visual file. If that works reliably, it could help teams search not only for text but for relationships: the dimensions of a room, the type of a wall assembly, the location of a specified component or the revision history attached to a zone.
What “reading” a blueprint should mean
The phrase “AI reads drawings” can be misleading. A system may recognise objects, retrieve notes and flag potential inconsistencies, but that does not mean it understands a project with the judgement of an architect or engineer.
A useful system should perform specific, testable actions.
It should be able to identify dimensions and labels. It should connect a door tag to its schedule entry. It should compare a requirement in a specification with a related object in the drawing. It should locate every appearance of a particular material or equipment type. It should distinguish between a current plan and a superseded issue.
Most importantly, it should show evidence.
If the platform flags a non-compliant condition, the user needs to know which drawing, which sheet, which revision, which note and which rule triggered the result. A red warning icon without traceable sources is not design intelligence. It is another coordination problem.
The ideal workflow is interactive. A project manager asks, “Show all doors on Level 2 with a fire-rating requirement that differs from the schedule.” The system returns a list of candidates, each linked to the exact locations and documents that need professional review.
That reduces search time while keeping responsibility where it belongs: with the qualified people making decisions.
Searchdog’s 70% claim needs context
Searchdog’s claim of a 70% reduction in time spent analysing blueprint compliance is significant, but it should be treated as a company claim rather than an independently verified industry benchmark.
Review time varies sharply by project type. A straightforward commercial fit-out, a hospital, a high-rise, a semiconductor facility and a transport hub have radically different information burdens. A platform may save substantial time locating relevant files in one project, yet provide less value where requirements are ambiguous, drawings are incomplete or the coordination challenge depends on specialist judgment.
The most realistic productivity gain will likely come from repetitive information retrieval.
Consider a tender-stage review. A contractor needs to confirm whether a client requirement appears consistently across hundreds of sheets and documents. An AI-assisted system may scan and organise relevant references in minutes, creating a shortlist of areas that need deeper review. The contractor still needs to assess constructability, cost, sequence, risk and contractual meaning, but the first search layer becomes faster.
That is a meaningful improvement. It frees skilled professionals from hunting through files and lets them spend more time interpreting what they find.
Why this matters for AEC workflows
The AEC sector has spent years producing more digital information without always making it easier to use. BIM models have become richer. Reality capture has become cheaper. Cloud collaboration has expanded. Yet project teams still lose hours to document hunting, version confusion and manual checking.
Searchdog is targeting this operational gap.
The company has signed a supply contract with a major South Korean construction firm and is also working with a semiconductor company. Less than two years after its founding, it has been selected for Nvidia’s Inception startup programme and aims to enter the U.S. market in the second half of 2026.
That target matters because the U.S. construction market contains some of the world’s most complex documentation environments. Projects often involve multiple consultants, local codes, specialist systems, strict contractual requirements and fragmented technology stacks.
If Searchdog can operate across those conditions, it will face the true test of its platform: not whether it can process a clean set of drawings, but whether it can work within the messy reality of active projects.
The real battleground is trust
AI-assisted design review will succeed or fail on trust.
A project team needs to trust that the system is using the correct drawing issue. It needs confidence that it has captured the relevant note, dimension or specification. It must understand what the system cannot determine. It must retain a clear record of who reviewed, approved or rejected a finding.
This means design-review tools should be built around traceability rather than automated certainty.
The most valuable platforms will not claim to replace architects, BIM managers, code consultants or contractors. They will make these professionals faster by giving them a reliable route through growing volumes of project information.
For AEC firms, the question is practical: can the platform reduce low-value searching without creating new quality-control risks?
A good pilot should test that question with a defined use case. For example, a team could use the system to compare door schedules against plans, search for accessibility-related notes or identify material references across a tender package. The findings should then be checked against manual review, with results measured for accuracy, speed and missed issues.
Search is becoming a design tool
Searchdog’s proposition is powerful because it focuses on a problem every project team understands: finding the right information before an overlooked detail becomes a costly issue.
Its reported 70% review-time reduction remains a claim that will need validation across real project types. Still, the direction is clear. As drawings, models and specifications become more data-rich, the ability to search them intelligently will become a core design and construction capability.
The future of blueprint review may not be a machine replacing human expertise. It may be a better partnership: a system that finds the relevant evidence, connects the documents and directs attention to the places where judgment is most needed.
A drawing will remain a drawing. But it may finally become a database that project teams can actually use.
