AI Vastu Floor Plan — What the Software Can and Cannot Do, Honestly

Editorial commentary, not professional advice. Software does not replace a licensed architect, a structural engineer, or a Vastu pandit who knows your family.

The phrase "AI Vastu floor plan" sits at an awkward intersection. On one side is software that is genuinely good at constraint-solving: it can fit rooms onto a plot, respect setbacks, place doors so they do not collide with walls, and check directional rules at speed. On the other side is a body of classical knowledge — Vāstu Śāstra — that includes orientation, ritual context, owner-specific factors, and centuries of architectural judgement that is not trivially encoded.

This piece is about what the first side actually can do, and where it should stop.

What an AI floor-plan engine does well

An AI engine that has been built honestly should be good at five things.

1. Constraint-respecting layout

Given a plot, road width, jurisdiction, and a desired BHK, an engine can lay out rooms inside the buildable rectangle, respect local setbacks, keep service rooms adjacent to the kitchen, route a circulation spine, and avoid the obvious geometric mistakes — toilet over puja, bedroom door opening into kitchen, stairs in the north-east corner.

We have written about the geometric side of this in Setbacks: How Margin Becomes Architecture. The engine's job is to keep that geometry coherent across hundreds of corner cases.

2. Vāstu Puruṣa Maṇḍala overlay

A 64-pada Manduka or 81-pada Paramaśāyika grid is a deterministic overlay once the plot's facing is established. A good engine can drop the grid onto the plan and then attribute each major room to its zone — Agni in the south-east, Īśāna in the north-east, Yama in the south, and so on.

This is also where it can flag the Brahmasthāna — the central nine padas in the Manduka grid — and warn if a heavy load (toilet, stair, beam pillar) is about to land there.

3. Directional rule checking

Once the maṇḍala is in place, the room-direction rules from Mayamata, Manasara, and Brihat Samhita become a finite checklist: kitchen in south-east, master bedroom in south-west, puja in north-east, toilets away from north-east, staircase clockwise rising. An engine can run that checklist in milliseconds and produce a structured score per room.

We do this in our own Vastu report. It is one of the easier wins for software.

4. Variant exploration

Where humans get tired, the engine does not. If the kitchen-in-south-east constraint conflicts with the staircase-clockwise constraint on a particular plot, the engine can try ten variants — kitchen on the south-east wall, kitchen as an attached extension, kitchen with a buffer corridor — and report tradeoffs. This is where AI adds real value over a single architect's first draft.

5. Source citation

A well-built engine should attach a source to every finding. "Kitchen south-east — Mayamata Ch. 26" is a different kind of output than "Kitchen south-east." The first invites verification. The second invites trust without evidence.

We have made the design decision that every line of our Vastu report should carry a source — Mayamata, Manasara, Brihat Samhita, Atharva Veda, or Shulba Sūtra. If the source is silent on a particular point, the report says so.

Where AI should stop

Five places where AI Vastu floor plans should hand off to a human.

1. The owner-specific factors

Āyādi Ṣaḍvarga — yoni, vyaya, ṛkṣa, and the rest — depends on the owner's nakshatra and on the perimeter measurement. Software can do the math. It cannot judge whether the family is ready to accept a result that conflicts with their existing plot. That is a conversation, not a calculation.

We feed our Vedic birth chart tool into Āyādi rather than ignoring it, but the final say belongs to a competent human advisor.

2. Plot-specific irregularities

Real plots are not rectangles. They have road tapers, slope, water tables, neighbour structures, and survey errors. A good engine assumes a clean rectangle and warns where assumptions break. A licensed architect adjusts.

3. Sanction and structural drawings

Our drafts are conceptual layouts. A municipality will not accept them as sanction drawings. A contractor will not pour foundation off them. Both of those require a licensed architect's stamp and a structural engineer's calculations. We say this in every output and in our products page.

4. Ritual context

When and how to perform gṛha-praveśa, foundation laying (śilānyāsa), or vāstu-pūjā is not a layout decision. The Brihat Samhita timing layer can produce candidate dates — that is what our Karaṇa checker does — but the ritual itself belongs to a pandit.

5. Aesthetic judgement

Whether a particular elevation feels like a home is not in the rule set. AI can produce a rule-respecting plan that is also visually flat. A human architect closes the loop.

How to read an AI Vastu output without overclaim

Three habits that will save you grief.

Read the disclaimers, not just the score. A high "Vastu score" with no disclaimers is a sales sheet. A medium score with explicit tradeoffs and source citations is a usable design draft.

Cross-check the orientation. Is the engine using true north or magnetic north? Has the local declination been applied? A 6° error in orientation will move the entire maṇḍala by 6° — that can swing a kitchen out of agni and into nāirṛti.

Treat the layout as a draft. Even our most careful output is a starting point. Take it to a licensed architect. Take it to your family Vastu advisor. Sit with it for a week before you commit.

What we are deliberately not claiming

We do not claim:

• that our engine produces "perfect Vastu"
• that it replaces a licensed architect or structural engineer
• that an AI score above any threshold means the home will be auspicious
• that any single classical text alone is sufficient
• that owner-specific factors can be ignored

We do claim that we run the constraint-solving and the directional rule-checking against catalogued classical sources, that we declare the method, and that we surface tradeoffs in the output. That is the bar we hold ourselves to.

How this shapes a GrehYug-generated plan

When we generate a house plan draft, the engine does the layout work, runs the maṇḍala overlay, scores the directional rules with sources, and then writes a one-page summary of what it could not resolve. The summary is meant to be read in front of an architect, not used as a sanction drawing.

If your provider does not work this way, ask them why.

Sources used in this article

• Mayamata Ch. 6 (orientation), Ch. 7 (maṇḍala), Ch. 26 (residential proportion)
• Manasara residential and yoni chapters
• Brihat Samhita Ch. 53 (residential), Ch. 98–100 (muhūrta)
• Atharva Veda 3.12, 9.3 — Śālā Sūkta
• Baudhāyana Shulba Sūtra 1.12
• GrehYug internal KB on classical-source attribution per finding