50+ products — from AI-native startups to the software you already use — showing what each one can do, where it fits in a project, and the prompts that connect them.
A cross-tool workflow Archi Automate lists on its own feature page, run through MCP.
Generative AI went mainstream in 2022. Midjourney opened its public beta in July, Stable Diffusion was released as an open model in August, and ChatGPT launched in November. Architecture felt it first, because the first models made images — and within months, mood boards and concept renders were being generated instead of drawn.
Four years later the story has moved from pictures to production. AI now reads drawing sets and code books, generates floor plans and massing against zoning rules, writes Revit and Archicad scripts from a sentence, and increasingly works inside the model itself. Engineering is following with AI that routes MEP systems and calculates embodied carbon.
And construction now has its own wave: progress tracking from 360° site captures, schedule optimization across millions of scenarios, automated takeoffs and cost estimates from plans, and AI drawing review that flags issues before an RFI is written. This map follows all three — the AI-native tools, the established software that has added AI, and the prompts starting to connect them.
Every product with AI in our catalog, sorted by the people who use it. In each column, the first group are new tools built around AI from day one; the second is software you may already own that now ships AI features. Tools that serve more than one discipline, like Autodesk Forma, show up in each column they serve — click any name to learn more.
Each point is one product, placed on the same six parameters we use on our comparison page. Change the axes to explore any two of them — ease of learning, BIM integration, collaboration, affordability and more. Colour and shape show who it's built for; click any point to open it.
Scores run 0–10 and are the same ones shown on each product's page and in our comparisons; higher is always better — a higher Learning Curve score means easier to learn, and a higher Price score means more affordable. BIM integration averages IFC/openBIM and Plugins. These are editorial estimates, not vendor-verified figures.
Every tool on this page uses AI, but in very different ways. Knowing which kind of AI is at work helps you understand what each one is designed to do — and where it can help your project.
Diffusion models trained on huge image sets turn a prompt, sketch or screenshot into a new picture.
Great at: mood, materials, lighting, fast client visuals.
Watch out: the output is pixels, not geometry — it can quietly change your design.
Rules, optimization and machine learning produce many layout or massing options against your constraints, then score them.
Great at: unit mix, parking, yield and floor-plan options in minutes.
Watch out: results are only as good as the constraints — most tools are tuned for housing.
Zoning codes, planning overlays and environmental analysis turned into instant checks on a real site.
Great at: ruling a site in or out early, envelopes, sun and wind feedback.
Watch out: coverage depends on the jurisdiction — confirm your city is in the dataset.
Computer vision that recognises walls, doors, symbols and annotations in DWGs, PDFs and scans.
Great at: legacy 2D sets into BIM, and scanning drawing sets for errors.
Watch out: messy or low-resolution drawings still need a human pass.
Large language models that read codes and documents, answer questions and take actions inside Revit or Archicad.
Great at: code research, documentation, scripting and repetitive modeling tasks.
Watch out: they can sound confident and be wrong — check citations and every change.
The first wave of AI tools made images. The newer wave works on the model itself — reading drawings, generating native geometry and running tasks inside Revit and Archicad. That's the shift to watch: AI that works where your project data already lives.
The most useful AI skill in AEC right now isn't picking a tool — it's writing the instruction. A good prompt can research a code, write a Revit script, or move data between programs that were never designed to talk to each other.
MCP (Model Context Protocol) is an open standard, published by Anthropic in 2024, that lets an AI assistant — Claude, ChatGPT, Gemini — call tools inside other software through a "connector".
Once your BIM software has a connector, you stop copying data between windows: the assistant reads the model, runs the command and reports back. The same prompt can pull a site from QGIS, build massing in Revit and send it to Blender.
The safe way to start: ask read-only questions first ("list", "count", "check") and only let the assistant change the model after you've reviewed what it plans to do.
Most tools on this page are polished products built on AI models you can also reach directly. You don't need any of this to use them — but if you're the technical one in your office, these are the places worth knowing.
Hugging Face is where the open AI community shares its work: more than 3 million models, plus the datasets they were trained on and Spaces — which it calls "the AI App Directory" — where anyone can try a model in the browser.
More and more of it is about buildings. Search for "floor plan" and you'll find 183 models, 144 datasets and 112 Spaces. In 2023 only 15 of them were published; in 2025 it was 198 — more than thirteen times as many. Interior design shows the same pattern: Spaces for it went from 7 in 2023 to 50 in 2025.
Why it matters for the industry: every shared model or dataset is a head start for the next one. A startup, a research lab or a firm's own computational team can build on a floor-plan dataset someone else labelled instead of starting from zero — so ideas move from research demo to real product faster, and more of the tools on this page will keep appearing. It also means firms don't have to wait for a vendor: open models can be tested, adapted and run on their own machines, with their own project data.
The catch: it's a lab, not a product. Many community demos are experiments — search Spaces for "interior design" and plenty were sleeping or showing a runtime error when we checked. Great for exploring what's coming; not something to hang a client deadline on.
The general-purpose assistants from OpenAI, Anthropic and Google are the AI most architects and engineers already use every day.
Open-weight image models such as Stable Diffusion and FLUX can run on your own GPU through free, node-based tools like ComfyUI.
Add ControlNet — a technique that makes the model follow the edges, depth or line work of your own drawing — and the image keeps your building's geometry instead of inventing a new one.
More setup, more control, no per-image credits — and when it runs locally, your project files never leave the office.
Hugging Face figures from huggingface.co on 26 September 2026: total model count from the Models hub; "floor plan" counts are models, datasets and Spaces returned by a search for "floor plan", grouped by the year each was created (2026 covers January to September).
Most AI tools look brilliant in a two-minute demo. These are the questions that decide whether one survives contact with a real project.
MCP (Model Context Protocol) is an open standard, published by Anthropic in 2024, for connecting AI assistants to other software. When a BIM tool has an MCP connector, an assistant like Claude or ChatGPT can query the model and run commands in it from a plain-English prompt. In our catalog, Archi Automate and Catenda Hub describe themselves as MCP-based, and NonicaTab offers an AI Connector for Claude Desktop inside Revit.
Buildots tracks site progress against the plan from 360° captures, ALICE Technologies optimizes construction schedules, Togal.AI, MeltPlan and zzTakeoff use AI for takeoffs, PLAN0 turns plans into cost estimates, and Helonic and Bluebeam Revu offer AI drawing review.
Several run directly inside Revit rather than as a separate app: Finch 3D for generative floor plans, WiseBIM AI for 2D-to-BIM conversion, Swapp and Ideatura for documentation, Augmenta for electrical system design, and BIBIM, Archi Automate and NonicaTab for AI-driven automation. D5 Render connects through a plugin too.
It depends entirely on the vendor and the plan. Some tools process your files in the cloud, some run locally inside your BIM software, and terms on training data vary — enterprise plans often offer stronger guarantees. Check each vendor's terms and privacy policy before uploading client work, especially if your contracts include confidentiality clauses.
Spacio, Archistar and TestFit all target this stage, with slightly different focus: Spacio and Archistar lean toward zoning/code compliance and site envelopes, while TestFit is built specifically for parking layouts and unit mix on multifamily and mixed-use sites.
Maket, Finch 3D, ArkDesign and Hypar all generate floor plan options from a set of constraints, but they're not interchangeable: Finch 3D lives inside Rhino, Grasshopper and Revit as a plugin, Hypar is a broader cloud platform, and Maket and ArkDesign are standalone tools focused on residential and multifamily layouts respectively.
It depends on your starting point. MyArchitectAI and Veras turn CAD/BIM models or sketches into photoreal images directly. D5 Render is a full real-time rendering engine with AI features built in, closer to Enscape or Lumion than a pure AI plugin. Magnific isn't a renderer at all — it upscales and sharpens a render you already have.
Yes, within limits. WiseBIM AI is a Revit add-in built to detect walls, doors, windows and slabs from an imported 2D drawing (DWG, DXF, PDF or image) and generate the Revit geometry automatically, and QikBIM converts 2D drawings into coordinated models.
Several have a free or low-cost entry point: D5 Render has a permanent free Community plan, Spacio offers a free tier, and many of the paid tools offer a trial. Pricing and trial availability are listed on each tool's own review page and change often, so treat the numbers on this page as a starting point.
We'll keep this map updated as pricing changes and new tools prove themselves — if you use one we haven't covered here, let us know.
Each product on this page has its own page in our directory, with features and links to the vendor.
Pricing and feature claims were verified directly on each vendor's own site in September 2026 and may change — always confirm current pricing before buying.