Using AI to Write Architecture Project Descriptions & Bids

Every architect has stared at a blank page at 11pm trying to write a project description that sounds professional without being boring. Or worse, grinding through a 40-page PQQ response where 30 pages are boilerplate you've rewritten slightly differently for the fifteenth time this year.

Writing is the unglamorous backbone of practice. Bid submissions, design and access statements, project descriptions for your website, fee proposals, planning application narratives -- they eat hours that could be spent designing. And here's the uncomfortable truth: most architects aren't trained writers. We're trained designers who happen to need words to get projects built.

AI writing tools -- ChatGPT, Claude, Gemini, and others -- won't make you a great writer. But they'll get you from blank page to solid first draft in a fraction of the time. The trick is knowing what to automate, what to edit heavily, and what to always write yourself.

Where AI Writing Actually Helps Architects

Not all architectural writing benefits equally from AI. Here's where the time savings are real.

Bid and Tender Responses

Most bid documents follow predictable structures: company profile, relevant experience, methodology, team CVs, project-specific approach. AI excels at generating structured responses from bullet points.

You provide: "We completed a 200-unit residential scheme in Manchester, delivered under budget, BREEAM Excellent, involved community consultation, completed 2024." AI produces a polished paragraph with proper structure, active verbs, and quantified achievements.

Time savings: A typical PQQ response that takes 8-12 hours can drop to 3-5 hours with AI-generated first drafts. You're editing and refining rather than staring at empty cells.

Design and Access Statements

D&A statements for planning applications follow a formulaic structure: site context, design rationale, access strategy, sustainability approach. AI can generate each section from your design notes and site analysis, producing a first draft that hits the expected structure and vocabulary.

Time savings: 4-6 hours of writing can compress to 1-2 hours of editing.

Project Descriptions (Portfolio and Website)

The 150-250 word project blurb that needs to sound compelling without overselling. AI is surprisingly good at this because it's trained on thousands of architecture studio websites. Give it the facts -- project type, size, location, materials, design concept, completion date -- and it'll produce something publishable with minor edits.

Fee Proposals

The narrative sections of fee proposals ("our approach to this project") are another natural fit. AI can structure your methodology, articulate your value proposition, and present your team's experience in a format that reads like a well-organized proposal rather than a brain dump.

Client Reports and Updates

Monthly progress reports, design stage summaries, and consultant briefing notes follow predictable templates. AI can draft these from your bullet-point notes, saving 30-60 minutes per report.

Where AI Writing Falls Flat

Genuine Design Thinking

If you ask AI to explain why your building curves along the riverfront or why the material palette transitions from brick at the base to zinc cladding above, it'll produce something generic about "responding to context" and "creating visual interest." These are the paragraphs that make architecture writing boring. Your actual design reasoning -- the conversation you had on site, the reference to a building in Porto, the client's unexpected brief requirement -- is what makes a project description compelling. AI can't access that.

Tone and Voice

Every practice has a voice. Some are academic and precise. Some are casual and confident. Some are poetic. AI defaults to a mid-range corporate tone that sounds like every other practice. If your studio has a distinct voice, you'll need to edit heavily or train the AI with examples of your existing writing.

Technical Accuracy

AI confidently generates plausible-sounding technical claims that may be wrong. It might describe a "post-tensioned concrete frame" when your building has a steel structure, or reference "Part L compliance" when the project is in Scotland (where it's Section 6 of the Building Standards). Always verify technical statements. AI doesn't know your project -- it's guessing based on patterns.

Anything Confidential

Don't paste client briefs, financial information, or unpublished project details into public AI tools. Most platforms state in their terms of service that inputs may be used for training. Use enterprise versions (ChatGPT Enterprise, Claude for Work) if you need confidentiality, or run local models.

Prompt Templates That Work

The quality of AI output depends entirely on your prompt. Vague inputs ("write a project description") produce vague output. Structured prompts with specific information produce drafts worth editing.

Project Description Prompt

Write a 200-word project description for an architecture studio website.

Project: [Name]
Type: [Residential / Commercial / Mixed-Use / Cultural / etc.]
Size: [GFA in sqm or sqft]
Location: [City, Country]
Completion: [Year]
Client: [Name or type, e.g., "private developer"]
Key design features: [3-5 bullet points]
Materials: [Primary materials]
Awards/certifications: [If any]
Tone: [Professional / conversational / academic]

Focus on the design concept and user experience. Avoid cliches like
"seamlessly blends" or "redefines the boundaries." Use active voice.

Bid Methodology Prompt

Write a 500-word methodology section for a bid response.

Project type: [Brief description]
Key challenges: [3-4 bullet points]
Our approach: [3-4 bullet points on how we'd tackle it]
Team structure: [Key roles]
Software/tools: [BIM, sustainability tools, etc.]
Relevant experience: [1-2 similar past projects, one line each]

Structure as: Introduction (2 sentences), Design Approach (paragraph),
Technical Delivery (paragraph), Collaboration & Communication (paragraph),
Programme & Risk Management (paragraph). Professional tone, no jargon.

Design Statement Prompt

Write a 300-word design statement for a planning application.

Site: [Address, context description]
Proposal: [What we're building]
Design rationale: [Why it looks/works the way it does - 4-5 bullet points]
Context response: [How it relates to surroundings]
Sustainability: [Key strategy, e.g., Passivhaus, BREEAM target]
Access: [Key access considerations]

Write in third person ("the proposed development..."). Formal but
readable. Reference local plan policy names if provided. Avoid
subjective claims about beauty or aesthetics.

Fee Proposal Approach Section

Write a 400-word "our approach" section for a fee proposal.

Project: [Brief description]
Client priorities: [What the client cares about - cost, speed, quality, sustainability]
Our differentiators: [What we offer that others don't - 3 points]
Methodology: [How we'd deliver - key phases]
Team: [Key personnel and their relevance]

Tone: confident but not arrogant. Focus on client benefits, not
self-promotion. Use "we" not "the practice."

The Editing Process: What to Always Fix

AI drafts are first drafts. They need editing before they're client-ready. Here's a checklist.

Always fix:

• Generic phrases. Replace "state-of-the-art" with specific descriptions. Replace "thoughtfully designed" with what you actually thought about.
• Passive voice. AI loves "the building was designed to..." -- change to "we designed the building to..."
• Unsupported claims. If AI says "the project achieved a 40% reduction in energy use," verify the number or remove it. Don't publish AI-invented statistics.
• Wrong technical terms. Check building regulation references, material specifications, and structural descriptions against your actual project.
• Missing specificity. AI writes "large windows provide natural light." You should write "floor-to-ceiling glazing on the south elevation delivers a daylight factor of 3.2% to the open-plan workspace."
• Repetitive structure. AI tends to start every paragraph with "[Noun] + [verb]." Vary your sentence structure.

Always write yourself:

• The opening sentence of any design statement or project description. This is where your voice matters most.
• Any paragraph explaining your design concept or philosophy. This is your thinking, not a template.
• Client-specific content that references conversations, site visits, or brief requirements. AI can't know what happened in your meetings.

Real Time Savings Data

Based on conversations with practices using AI writing tools in 2025-2026:

The savings are real but front-loaded. You spend less time writing but about the same time editing and checking. The net result is faster turnaround and more capacity to pursue additional bids -- which, for a small practice, can be the difference between feast and famine.

If you're hunting for your next project opportunity, ArchGee's architecture job listings include roles at practices of all sizes where bid writing and project communication are valued skills alongside design ability.

Ethical Considerations

Disclosure. Should you tell clients you used AI to draft their bid response? Opinions vary. Most practices treat AI as a drafting tool -- like spell-check or Grammarly -- that doesn't require disclosure. But if a client specifically asks, be honest. Misrepresenting AI-generated content as entirely hand-written would be misleading.

Quality responsibility. AI-generated text is your responsibility once you submit it. If the bid response contains an error -- a wrong project date, a fabricated statistic, an inaccurate technical claim -- that's your name on the document. AI doesn't have professional indemnity insurance. You do.

Intellectual property. Text generated by AI tools is generally not copyrightable (per current US Copyright Office guidance). If your project descriptions or design statements need IP protection, ensure they contain sufficient human-authored content to qualify.