For more than a decade, social platforms have had to deal with criticism of embedding algorithms that feed compulsive behavior of fuel, encourage doomscrolling and measure success over time. Pinterest, positioned for a long time as a calmer alternative, is now trying a little more brutal: reinvestion personalization by AI.
Chief Technology Officer Matt Madrigal calls this shift from ‘addictive to additive’, using algorithms as aids to help people design the digital worlds they want.
“In contrast to other platforms, the Pinterest algorithms are not designed to keep users addicted, which can lead to an overwhelmed, anxious and distracted,” he says Fast Company. “Our taste graph serves as the basis of our AI systems and makes our recommendations possible to deliver highly personalized experiences. It maps hundreds of billions of unique interactions that connect interests, goals and behavior to help users cut the noise.”
Madrigal, who once led the Merchant Products from Google Shopping, now guides the Foundation AI models from Pinterest with a philosophy that respects as much as relevance. “AI is at best when real human needs should instead of just optimizing for clicks or transactions,” he says. “During my time at Google I also saw how frictionless trade shoppers and brands could set. At Pinterest our AI strategy is not only about personalization for relevance, it is also about serendipity.”
The multimodal discovery system of Pinterest feeds visual searches that Madrigal says it is 30% more effective than leading ready-made models. Because of the inclusive AI functions, beauty and fashion search assignments can be refined by hair pattern, skin color and body type.
The philosophy also appears in product choices. Pinterest has added substantive labels generated by AI, a switch “fewer AI pins” and an option to refrain from having activity used to train models. “We reject the false dilemma that maximizing AI innovation means sacrificing responsible development,” says Madrigal.
Although rivals are confronted with control of algorithmic damage, Pinterest guesses that transparency and user control can be a competitive country.
[Photo: Pinterest]The responsible approach in the AI weapon race
Meta encloses AI deeper into Instagram and Facebook; Tiktok sharpenes recommendations and expands the trade; YouTube relies on AI to stimulate sticker engagement. Pinterest’s rollout from 2025 takes a different tack and emphasizes the responsibility as much as personalization.
New tools include personalized background generation, which transforms ordinary product images into lifestyle-ready visuals that are tailored to the aesthetics of a user, and performance+, an automation suite that reduces friction for marketers by optimizing advertisements with fewer inputs.
Madrigal points out as an example of Prada’s Leather view campaign: a decrease in costs per action by 64% and a conversion rate of 30%. He says that these advertising tools have contributed to a turnover increase of 19% on an annual basis to more than $ 3 billion.
The most visible change is the AI-generated content label. Pinnen made or processed with generative AI now wear clear tags, embodies what Madrigal calls ‘AI with guardrails’.
“There is a question from users and makers for more transparency and control over the generative AI content they see,” he explains. “AI is an important role in improving both user involvement and income income. Ultimately, embedding positivity within our business model has proved successful.”
[Photo: Pinterest]Can the AI involvement and responsibility balance?
Pinterest’s Last income Mark both traction and tension. In Q2 2025, sales reached $ 998 million and all -month -old active users up to 578 million, largely driven by international growth and the acceptance of Gen Z. Nevertheless, Wall Street shrugged – appears fell more than 12%Sliding in the range of $ 35 – $ 36, even if the turnover was at the top. Since November 2024, the company has also purchased nearly 10 million shares.
Commercial pressure, experts warn, often give in favor of stickiness above restraint.
“History has shown that listed companies are often forced to optimize for shareholder value, and for shareholder value they have to earn the most income and the most profit. This correlates directly with involvement, so they create loops that get the most engagement,” Ajit Varma, VP of product for Firefox at Mozilla, says a fast company says. “But because there are choices and people prefer those choices, there will always be a future in which AI serves humanity versus serving the profit interests of companies.”
That tension exposes the limits of a playbook with advertising-first Playbook. Varma argues that users must be able to ‘vote’ their choices for systems that reflect their values, whether it is open source, transparent or safety first. “Ultimately, it is up to users to show that they value these alternatives by choosing and using products that are built and use on different models.”
The challenge of Pinterest does not only produce that vision; It proves users and investors that additive AI can match addictive AI. If the PlayBook succeeds from 2025, it can create a precedent that lies real power in generative systems that improve creativity with respect for boundaries.
“I imagine a platform that not only understands what users are looking for, but also the context behind their interests, ambitions and moods. Ai will be the engine, but inspiration and positivity will remain our North Star,” says Madrigal. “In different teams we work on creating new AI solutions that put the entire end-to-end user experience in power and we bring in top talent to make it happen.”
#Pinterest #bets #Additive #personalization #represents