Technical & Structure · 9 min read · July 15, 2026
Your image-heavy portfolio is invisible to AI: how to make every project readable
Your portfolio is a gallery: renderings, floor plans, detail shots, barely a sentence of text. That works on people. It does almost nothing for AI. ChatGPT, Perplexity and Google AI Overviews read language, not aesthetics. If a prospective client asks who designs sustainable timber buildings in the region, you only surface when your projects are described in words and structured so a machine can parse them.
Why your best project doesn't exist for AI
Architecture firms invest heavily in visualization: high-resolution renderings, elaborate photo series, polished slideshows. That wins people over instantly. But generative AI systems like ChatGPT, Perplexity, Claude and Google AI Overviews work primarily with text. An image with no descriptive text is a blank field to them. If a project page is just a rendering titled Residence H., no AI system knows it's a passive-house timber-frame new build of 180 square meters in Freiburg.
The result: a prospective client asks their AI which architecture firm in Freiburg has experience with timber passive houses, and your perfectly matched project is never mentioned. The information simply doesn't exist in text form. A competitor with a plainly described, even visually weaker project gets recommended instead, because the AI can read what that project actually is. Beauty wins over people. Language wins over machines.
This is the core tension for your firm: the visual strength that sets you apart is exactly what makes you invisible to the newest generation of search. GEO, short for Generative Engine Optimization, doesn't undo that logic, it completes it. You keep your images and give the AI the words it needs to understand your work and recommend it.
What clients actually type into the AI
People phrase questions to AI systems differently than they do to a classic search engine. Instead of Architect Munich, they describe a whole situation: We want to renovate for energy efficiency and add a story to an existing 1960s building, which firm in Upper Bavaria has done that? Or: Who designs barrier-free multigenerational housing with KfW funding? These questions carry the building type, the region, and the constraints. Your website needs those exact terms in natural language, or the AI has nothing to connect to.
Collect the real questions from your initial consultations. Clients ask about cost per square meter, construction timelines, how the service phases unfold, experience with the heritage authority, dealing with sloped sites. Each recurring question is a topic that deserves its own clearly written passage on your site. AI systems draw their answers from exactly that kind of concrete text, then cite the source.
Think about the language of use cases too: daycare center, medical practice, winery, terraced-house development, commercial hall, attic conversion. If your references are all labeled Project 2023, you throw away every one of those anchors. Name what the project actually is. Not New Build W., but Construction of a two-group daycare center in timber, Regensburg. That's a sentence an AI can quote.
Every image needs its words: alt text and captions
The fastest lever is sitting right in your images. Every photo and rendering has an alt attribute in the source code that's almost always empty, or reads something like DSC_4821. That's exactly where a precise description belongs: Street-facing facade of the renovated Gründerzeit building with new clinker brick cladding and floor-to-ceiling windows, project in Leipzig. That text is written for screen readers, but AI systems and search engines read it just as closely.
Add visible captions under your project photos too. They serve people and machines equally well. Don't write Interior — write Open living and dining area with exposed-concrete ceiling and oak parquet flooring, 2.80-meter ceiling height. Captions like that turn a silent gallery into a narrated walkthrough of your project that an AI can absorb paragraph by paragraph.
What matters is honesty, not keyword stuffing. Describe what's actually visible in the image, using the technical terms you already use day to day. Materials, construction method, location and use are plenty. A string of search terms reads as spam to an AI and can drag the page down instead of helping it. Concrete, accurate description is always the stronger move.
The project fact sheet: the details AI can work with
Give every project a structured fact sheet in plain text, visible right on the page. Cover the details people actually ask about: location, building type, use, gross floor area, construction method, year built, service phases delivered, client (where you're allowed to name them), and any special factors like funding, energy standard or certification. This block is invaluable to an AI, because it can answer a question directly instead of guessing.
For example: location Ravensburg, project new construction of a mixed residential and commercial building, six residential units, 620 square meters gross floor area, solid timber construction, KfW Efficiency House 40, service phases 1 through 9, completed 2024. From that single block, an AI can answer several different questions: about timber construction, about Efficiency House 40, about mixed use, about the region. One well-structured project can cover a whole range of queries.
Keep the format consistent across every project. When each reference follows the same fact-sheet structure, an AI system recognizes the pattern and can compare your projects cleanly against each other. A human visitor barely notices the repetition, but to a machine it reads as a strong signal of reliability and completeness.
Structured data: the format machines read instantly
Beyond the visible text, there's Schema.org markup — an invisible data layer in your site's source code. For architecture firms, ProfessionalService or LocalBusiness markup, along with Project and ImageObject markup, matter most. It tells the machine explicitly: this is a firm, operating in this city, here are its projects, here are its services. A web developer can add it in a few hours. Note that this markup helps machines parse your site correctly — it isn't a guaranteed ranking or citation boost on its own.
Structured markup for your contact and location data pays off especially. When name, address, service area and specialties are stored in machine-readable form, an AI can reliably place you in a region and a specialty. That's exactly what determines whether you show up for a question like a firm near me. Without that data, your location stays fuzzy to the machine.
Add FAQ markup for your most common questions. When you turn the questions from your initial consultations into an FAQ with schema markup, AI systems are more likely to pick up those question-and-answer pairs and quote them directly. That's one of the more reliable ways to get named in a generated answer at all.
From picture book to written record: the project narrative
Give every important reference a real written narrative of two to four paragraphs. Describe the starting situation, the client's brief, your design response, and the outcome. Explain why you chose timber over concrete, how you handled a tight site, what funding you secured. This narrative is exactly the material AI-generated answers are built from.
These narratives do double duty: they convince the human reader who wants to know whether you understand their building task, and they hand the machine context and technical vocabulary. Write in your own technical language, but keep it readable. A sentence like The added story was built in lightweight construction so as not to overload the existing structure is valuable and quotable for both audiences.
Compare the two approaches directly: a pure image series tells an AI nothing about how you think. A project narrative makes your methodology, your priorities and your experience explicit — exactly what a client is looking for when choosing a firm for a demanding project. Text is the bridge between what you can do and the AI that's deciding whether to recommend you.
Reputation beyond your own website
AI systems base their recommendations on more than your own site — they draw on the whole web. Mentions in trade publications, architect directories, chamber listings, competition documentation and construction blogs strengthen your profile considerably. When a project is written up in a professional article and credited to your firm, an AI gains confidence and names you more often. You can build these external signals deliberately.
Keep your entries in Baunetz, chamber-of-architects directories and relevant portals as clearly written as your own site, and make sure your firm name, location and specialties are worded identically everywhere. Inconsistent details — sometimes Munich, sometimes Munich-Schwabing, sometimes a slightly different firm name — confuse the machine and weaken the match to you. Consistency across sources matters.
Awards and competitions are worth pursuing for this reason too. A prize is typically covered by several outlets, creating multiple consistent mentions of your name tied to a specific project. For AI systems, that kind of agreement across sources is a strong quality signal that pushes you higher in recommendations.
Concrete first steps for your firm
Start small and targeted rather than overhauling everything at once. Pick your five strongest reference projects and fully build them out: precise alt text on every image, visible captions, a structured fact sheet, and a short project narrative. Five well-prepared projects will do more for your AI visibility than fifty silent galleries. Quality of description beats quantity every time.
Then test it yourself. Ask ChatGPT, Perplexity and Google AI Overviews exactly what a client might ask — say, about a firm for timber construction in your region. Do you come up? Does a competitor get named instead of you? Look closely at how their page is described differently from yours. These tests will show you, in plain terms, where your gaps are.
Build the new standard into your process going forward. Every new project should get a fact sheet, image descriptions and a short narrative from day one, before it ever goes online. That way your AI visibility grows automatically with every commission, instead of staying a one-time cleanup. Step by step, an image-heavy portfolio becomes a body of work that machines can read and people can recommend.
Common questions
Do I have to replace my beautiful renderings with text?
No, not at all. Your images remain the emotional core of your portfolio and are what wins over human clients. You're only adding words alongside them: alt text, captions, project fact sheets and short narratives. Image and text work together — the image wins the person, the text is what makes you findable and quotable to AI systems in the first place.
For confidentiality reasons I can't name many project details. What can I do?
That's normal, especially with private clients. You don't need to name people or addresses. Describe the design-relevant facts instead: building type, general region, construction method, floor area, energy standard, the brief and your solution. Those factual details are what the AI needs, not the client's identity. An anonymized but technically precise fact sheet is fully effective and stays confidential.
How quickly will I see results after rebuilding my portfolio?
Classic search engines take a few weeks to re-index changes. AI systems vary: some pick up updated content fairly quickly, others only catch up at the next training or index refresh. Realistically, expect anywhere from a few weeks to a few months. Consistency is what matters — describing every new project properly and building up outside mentions leads to being recommended more often by AI over time.
Read on