An AI-enhanced rendering can make a student portfolio look finished in minutes. It can also make an employer wonder which parts of the project the applicant actually designed. That is the question your portfolio must answer. The choice is not simply “use AI” or “never use AI”. It is whether each page gives credible evidence of your own architectural thinking.
In a discussion among practicing architects about AI in applicants’ portfolios, opinions ranged from rejecting generated images to accepting a clearly labelled example that shows process. This is a discussion sample, not a survey of employers. Its useful lesson is that a glossy image can create opposing reactions. A process-led page gives both types of reader something concrete to evaluate.
Can you use AI in an architecture job portfolio?
Yes, provided the employer or school does not prohibit it, you have the right to use the source material, and the page accurately states what you did. Whether a particular firm likes AI-generated imagery is a separate hiring preference. The AIA’s hiring-manager advice emphasizes showing thought from brief through sketches, models and outcome. The RIBA portfolio guidance similarly stresses explaining the process and your contribution.
The practical rule: the more a generated output changes the design or final image, the more clearly you should explain its role. A spell-checker on a caption is a different kind of assistance from a tool that invents a facade, an interior layout or an entire site. There is no universal disclosure threshold; describe material assistance in plain language and follow the specific application’s rules.
What a hiring manager needs to see
An architecture portfolio is evidence for a job. It should let the reader answer five questions quickly:
- What was the brief and who was the work for?
- Which part did you personally create or decide?
- Which constraints shaped the work?
- How did the proposal change after testing or feedback?
- What can you do again in a new office?
An AI image by itself answers almost none of these. A hand sketch by itself may not answer them either. Pair the image with a plan, a diagram, a detail, a test or a short account of a decision. In another portfolio review by architects, reviewers asked for technical drawings behind beautiful visuals. Their feedback points to a common hiring problem: aesthetic finish can conceal the absence of buildable thinking. For an early-career example, see our guide to junior architecture work in the AI era.
A useful disclosure line: tool, source, decision, check
Put a short note next to the relevant image. For example:
“I modelled the massing and facade openings in Rhino. I used an image-generation tool to explore lighting and planting for this study view. I restored the original window geometry, checked the material palette against the model and produced the final layout. Academic concept; not a built project.”
The wording is an illustrative template, not a claim about your work. Edit it to match what happened. If an external visualiser, classmate or colleague also contributed, credit them. If the project belongs to a former employer, check permission before uploading any drawing, model or image to a third-party service. For the broader selection of work, see the jobs.archi architecture portfolio guide.
A good note is specific without turning a portfolio into a software log. “AI used” is too vague. “Designed entirely by me” is misleading if the tool supplied crucial design decisions. “Generated for visual exploration; geometry and performance unverified” gives the reader a fair frame for a speculative image.
Build one page that proves authorship
Try a five-part page instead of a full spread of final renders:
- Brief: one sentence naming the user need and site condition.
- Your starting material: a sketch, plan or model view made before AI processing.
- AI’s role: the exact task, such as testing atmosphere or comparing facade treatments.
- Your intervention: an annotated example of a generated error you corrected or a direction you rejected.
- Final evidence: a plan, section or detail that confirms the design shown is the design intended.
For an academic housing project, you might compare a generated courtyard visual with the actual section. If the image made the courtyard wider, mark it and replace the inaccurate view. That correction is more impressive than pretending the problem was never there. It shows you understand the difference between persuasion and representation.
Do not add an AI image merely to demonstrate that you know a tool. Include it when it improves the explanation of a decision. If it cannot be tied to a design question, it may consume valuable portfolio space.
Four failure modes to catch before you send a PDF
1. The image changes the architecture
Compare generated views with your model. Look at structural spans, wall openings, stairs, floor levels, accessibility routes, roofs and neighbouring buildings. Zoom in. An image that silently changes the design could make you appear careless in a review, even if a viewer never notices the original mistake.
2. The portfolio implies a building exists when it does not
Label concepts, competitions, student work, unbuilt proposals and completed buildings accurately. Do not use photorealistic atmosphere as a substitute for project status. If the image is speculative, say so near the image rather than hiding it in a final page of credits.
3. The prompt replaces your decision
If your only explanation is “I asked AI to design a library”, the employer cannot see your ability to analyse a brief or make a trade-off. Show the constraints you supplied and the options you rejected. Be ready to sketch or explain the scheme without the generated image in front of you.
4. The output contains material you cannot share
Client files, unpublished site information, a former firm’s model or another creator’s image may be restricted. A seemingly harmless upload can disclose source material to a service with terms you have not checked. Use self-owned academic work or permission-cleared material for experiments. The AIA AI Firm Toolkit advises treating unverified tools as public with respect to confidential information.
Prepare for the interview, not just the PDF
Pick one AI-assisted page and rehearse a two-minute explanation:
“The brief required X. I built Y. I used the tool to explore Z. It produced A and B, but A conflicted with the actual section. I rejected it, adjusted C and verified the final view against D.”
Replace those placeholders with your real project. An interviewer may ask why you accepted a generated option, how you checked it, or what you would do differently without the tool. If you cannot answer, the project is not ready for the portfolio.
Keep a non-AI example nearby. It need not be a masterpiece: a clear detail, site diagram or iterative plan can show fundamentals. In the Reddit hiring discussion, some reviewers welcomed a small, labelled AI experiment while objecting to an entire portfolio of generated imagery. A mixed portfolio lets the work speak for itself without requiring the reviewer to guess at your baseline skills.
A quick self-audit for every AI-assisted page
Before sending the file, ask:
- Could a reader identify my personal contribution in 15 seconds?
- Does the image match the plans and sections shown?
- Are project status and collaborators labelled accurately?
- Do I have rights to use both the input and the output here?
- Could I explain the design with the image hidden?
- Would a firm with a cautious AI policy still see a useful architecture case?
If two answers are no, revise the page before changing the cover design. The easiest fix is often to remove an image and enlarge a drawing that reveals a decision.
Frequently asked questions
Should I tell an employer I used AI for post-production?
Disclose meaningful image changes and follow the employer’s instructions. A short note stating that AI was used for planting, lighting or cleanup is clearer than leaving the reader to infer it. If the tool modified geometry or materials, describe that and correct any mismatch before sending the portfolio.
Will an AI-enhanced render get my application rejected?
Some reviewers dislike it; others accept a well-explained example. There is no reliable universal rule. Your strongest defence is a project page that proves your own decisions, credits collaborators and makes clear what was generated. Do not build an entire application around a single visual effect.
Can I use AI to improve work from a previous firm?
Only after checking your rights, confidentiality obligations and the tool’s handling of uploaded data. A portfolio permission is not automatically permission to upload source files to an external model. When in doubt, build a new self-initiated exercise with material you own.
The takeaway
The best AI-assisted portfolio page is not the most seamless image. It is the one where a hiring manager can trace the brief, your input, the tool’s contribution and your verification. Show a design decision that survived scrutiny. That is evidence of the architect an employer would be hiring.
The post AI in Your Architecture Portfolio: Show the Work Behind the Image first appeared on jobs.archi.