AI Changed Architecture. Is It Changing How You Judge It?

Polish is supposed to make a submission easier to judge. But as our recent webinar made clear, that's not how it's playing out for AIA programs in practice — when AI can turn a two-to-three-week rendering into a two-to-three-day one, polish stops telling a jury anything about the design underneath it, and the burden of telling the two apart lands squarely on the people judging it.

In AI Changed Architecture. Is It Changing How You Judge It?, Reviewr walked through what an AIA judging process actually needs in order to score design fairly without banning AI outright. If you missed it live, here's what we covered, and you can catch the full session on demand below.

A recorded session on building a judging process that scores design, not polish.

Every submission that comes through an awards, scholarship, or competition program today has some form of AI behind it, and that isn't changing back. AI-assisted renderings have cut visualization time from roughly two to three weeks down to two to three days, so firms of every size can put a polished submission in front of a jury without ever needing a dedicated rendering team. One firm's shortlist rate climbed from 15% to 35% simply by submitting more AI-assisted concept variations — output volume alone is already reshaping outcomes. The AIA's own 16-member AI task force has named this directly: AI is raising new questions for the profession, and the question that matters most to a jury is how to score it.

The session also named the risk plainly: polished AI output often appears credible despite incomplete underlying assumptions, a documented pattern researchers call automation bias, where reviewers under deadline pressure over-trust a slick rendering. A second risk was flagged alongside it — homogenization. AI tools are trained on a limited set of references and quietly pull design toward familiar, trending styles, which is exactly the kind of distinctive design an awards program exists to recognize in the first place. The session's core message: the answer isn't banning AI, it's a rubric AI can't fool.

What We Covered

1. Set Your AI Expectations Up Front: Your policy and your acknowledgment go directly into the application form, not a PDF nobody reads. Using Reviewr's form editor, expectations go at the front of the submission, and applicants have to confirm they've read them before they're able to move through the rest of the form. That acknowledgment protects you as the admin — if someone is declined later for using too much AI, they already agreed to the terms on record. And because AI norms are still shifting, your policy isn't locked in: it can be updated inside Reviewr at any point, without setting aside an entire day to rebuild the form.
2. Treat AI Detection as a Data Point, Not a Verdict: Reviewr's built-in AI detection flags each submission so you can see how much AI was used against the threshold you set in your policy. Judges see the flag, know what's already been deemed acceptable, and weigh it alongside everything else in the submission rather than disqualifying an entry automatically the moment any AI use shows up. The result is a consistent, structured signal every cycle, instead of judges guessing case by case.
3. Let AI Summarize, Not Decide: This is a feature Reviewr rolled out in just the last few months, and client feedback on it has been strong. AI summarizes the parts of a submission you choose — not the whole application — giving jurors a fast, accurate first read on dense, rendering-heavy entries. That's what lets judges triage: does this move on to more intensive judging, or is it clearly not in the running for this round? The score and the final call always stay with your judges — the summary just means their real time goes to the entries that actually deserve a debate.
4. Score the Design, Not the Polish: Every judge works from one scorecard, customized to align exactly with your submission form, so there's no gray area where a judge is asked about something they can't actually find in the application. Questions are weighted, too — a more intricate design element that AI can't fake carries more weight than a quick AI-generated rendering, and Reviewr calculates the final weighted score automatically. The effect: a modest, well-reasoned entry can score as strongly as a highly rendered one, and standards stay consistent even as rendering quality keeps climbing.
5. Build a Process That Keeps Up: Your questions, requirements, criteria, and workflows can be edited in Reviewr at any point — even mid-cycle, if you need to add a question. Configurations, automations, and workflows from past years carry forward directly, so you're not starting from a blank form every cycle. That matters because AI, architecture, and the AIA world are all still evolving, and a judging process that's locked in place until it's rebuilt from scratch can't keep pace with any of it.

What the Live Demo Showed

• A branded microsite, linked from the AIA program's own website, carrying the program's colors and banner so applicants know they're in the right place
• Account creation for admins, applicants, and judges alike, with automatic save-progress as they move through the form
• Incomplete-submission reminder emails, a submission confirmation email, and a confirmation page once an application is in
• Every question type Reviewr supports, high-resolution image and file collection across 130+ supported file types, and essay prompts that give applicants a voice beyond the renderings
• A letter-of-recommendation process where references are invited directly into Reviewr to upload their own letters, straight from the source
• A judge portal showing every assigned evaluation with running score totals, so judges can catch and fix an outlier score up until the evaluation deadline
• A combined submission-and-evaluation view — every question, response, file upload, and supplemental form on one side, the aligned evaluation form on the other — instead of judges juggling a dozen open tabs and downloaded files

The Recap: Old Way vs. New Way

Old way: AI use policies are buried in a PDF nobody reads, so there's no real way to enforce them. AI-flagged entries are either treated as instant disqualifications or ignored completely. Jurors read every page of every submission cold, under deadline pressure, with no way to triage what actually deserves a closer look. Judges score from a personal read with no clear guidelines — this looks good, this doesn't — and the judging process itself is locked in place until it's rebuilt from scratch the next cycle.

New way: AI instructions and acknowledgment are built directly into the application and required before an applicant can move through the form. AI detection surfaces as context for the jury, never an automatic verdict. AI-assisted summaries give jurors a fast, accurate first read, so their time goes to the entries that need real debate. Every juror scores from the same structured, weighted criteria — and your criteria and workflows adapt just as fast as the tools your applicants are using.

About Reviewr

Reviewr is built for programs like this one. We've been doing this for over 15 years, running thousands of applications every year, all in one connected platform built for awards, scholarships, grants, and competitions of every kind. A Reviewr team member is with you every step of the way — one-on-one calls and hands-on assistance included — so getting your AIA program running the way you need it isn't something you're figuring out alone.