Content Benefit

Content Strategy | UX Writing | Content Design | Globalization | Communications

The AI is picking my sofa

The AI is picking my sofa

AI interior design: the search boom around Planner 5D and its rivals

Fall is the season for new curtains, freshly painted walls and, at long last, replacing that couch that has seen one too many evenings. This year, before reaching for the tape measure or the catalog, plenty of people open an app: searches for AI interior design, with Planner 5D in the lead, have grown by more than 1,000% and now add up to hundreds of thousands of queries a month.

The market is heading the same way. According to a report released on GlobeNewswire in July, AI applied to interior design was worth $1.39 billion in 2025 and is expected to reach $4.55 billion by 2030, a compound annual growth rate of 26.8%.

More than the figures, though, I’m curious about the question underneath: what happens to the “try before you buy” experience when the perfect sofa is suggested by an algorithm?

What Planner 5D, IKEA Kreativ and Wayfair Muse actually do

Planner 5D has been working with AI since 2016, when it started turning photographed floor plans into editable 3D models. In June it introduced agents dedicated to interior design, able to review a project, generate concepts and variations, retouch reference images and put together presentations. CEO Andrey Ustyugov argues that the next generation of these systems will run complete workflows instead of stopping at ideas.

IKEA took this road back in 2022 with Kreativ: you snap a photo of the room, the app rebuilds it in three dimensions with measurements, erases the existing furniture and lets you place items from the catalog. Wayfair chose another path. In 2025 it retired Decorify, which asked users to upload a picture and pick a preset style, and launched Muse, which creates rooms from a written description (“1940s living room with fireplace, patterned sofa and green rug”). Every scene is shoppable: suggested products are grouped by category and the buy button is always a tap away.

Three approaches for three kinds of users: those who want to draw, those who want to see their own living room transformed, and those looking for inspiration who end up finding a shopping cart.

Almost everyone ignored the augmented reality try-on

Before generative AI, the promise was called augmented reality: point your phone at the floor and the sofa appears right there, to scale. The Baymard Institute, which studies e-commerce usability, found that only 6% of its test participants used the “View in room” feature on their own, while 87% of those who came across it chose to skip it.

The reasons are very practical: disappointing past experiences, rooms too small or too cluttered to frame, vague instructions, imprecise gestures that resize the model without the user noticing, grainy 3D objects or wrong colors. When the purchase is a big one, people trust high-resolution photos, videos (41% watch them), shots uploaded by other customers and easy-to-find dimensions far more.

The lesson applies to AI too: an attractive simulation that gets the crucial details wrong ends up ignored. And a sofa is full of such details, as I wrote when exploring why design makes us crash on the couch every night.

ChatGPT has overtaken the specialized apps

Back in August 2025, Business of Home pointed out a paradox: the AI interior design startups born in 2023 had more or less stalled, while the most formidable competitor turned out to be ChatGPT, which according to an Elle Decor survey offers something close to a real consultation. It analyzes photos, generates moodboards and renders, recommends items and now lets you buy them too. With a well-known limit: every now and then it makes up products that don’t exist. Many of those young companies have fallen back on B2B, selling their tools to real estate agents and furniture retailers.

Google is also pushing Gemini for visualizing renovations. A Livingetc journalist, after a workshop at Google’s headquarters, acknowledged the tool’s usefulness but flagged its most annoying flaw: ask for a small tweak and you often get an entirely new image, and feeding earlier results back in makes the quality worse.

I asked Meta AI to rearrange my room

I wanted to try it myself with Meta AI, which offers free room restyling inside Instagram and WhatsApp. My request was simple and very concrete: rearrange my bedroom, adding a painting I actually own plus a side table and a lamp I’m about to receive as a gift. No impossible dreams, just real objects that had to live together. The result? Bland, flat, not the least bit appealing: the kind of picture you glance at for a second and forget right away.

To be fair to the machine, I have to admit that humans haven’t thrilled me either. The interior designer I consulted for my beach house didn’t put much imagination into it, and as for the IKEA planners, the less said the better. At this point, suspecting that the problem might be me, with a taste that’s hard to put into words, seems reasonable.

I’m not giving up, and I’ll try again with other AI solutions. In the meantime I take some comfort in a finding from the survey Planner 5D ran among its own users: one of the main difficulties they mention is precisely AI’s limited ability to grasp personal style.

Renovators use AI to imagine, then hire a professional

The Houzz study published in September 2026, covering more than 3,100 US homeowners in the middle of a renovation, describes a very specific kind of use. 22% have already tried AI, with a huge generational gap: 41% of millennials versus 7% of seniors. It’s mostly used to gather ideas (64%), visualize concepts (58%) and compare layouts, products and finishes (50%). When it comes to cost estimates the figure drops to 24%, and to 11% for scheduling. And 80% still hand the work to a professional; among those who don’t use AI, the most cited reason (38%) is a preference for human expertise.

Things are moving on the other side as well: 39% of design firms use AI, and 82% of them rely on it for early visuals. Planner 5D’s 2024 survey of 2,626 users reports 85.4% adoption, a figure to take with a grain of salt since the respondents were the platform’s own customers. Another number is more interesting: 74% would be willing to share their cultural background to get better suggestions, and almost 80% would like recommendations based on the emotions a room should evoke, such as calm, energy or inspiration. I’ve already explored the relationship between designers and machines in Designers and AI: who decides.

When the advice comes from whoever sells the sofa

This is the knot that interests me most as a UX writer. Kreativ and Muse are retailer tools: the “perfect” sofa will always be one they have in stock. That’s legitimate, but it changes the nature of the suggestion, which stops being a neutral consultation and turns into a tailor-made shop window. ChatGPT draws from several stores, yet remains opaque about how it chooses what to show.

Heavy personalization also carries a cognitive cost I know well: I wrote about it in Personalization is tiring, arguing that an interface stays manageable only when it’s designed around personas. And if tomorrow an AI agent can also order the couch on our behalf, the issue of trust when AI acts on its own comes right back.

What the try-before-you-buy UX needs

Putting together tests, surveys and my own small frustration, I see five priorities for anyone designing these tools:

  • Reliable measurements before a pretty render. If the footprint is wrong, nothing else matters.
  • Consistent edits. Changing the color of a cushion shouldn’t redraw the whole room.
  • Transparency about where suggestions come from. Say clearly whether they’re drawn from a single catalog or from the whole market.
  • A bridge between the generated image and the real object. Price, availability and dimensions of the actual product, right next to the dreamed-up one.
  • Room for the professional. Since 80% of homeowners hire one anyway, it pays to design the handoff instead of pretending it doesn’t exist.

Meanwhile, I keep looking, with AI and without it, for someone or something able to read my mind when I say “cozy but not too crowded.” I’ll keep you posted.

Related sources

Chiara Bonifazi