AI is making Meta's apps basically unusable, as the company adds Meta AI everywhere and viral user-generated AI images proliferate on Facebook and Instagram
partly because is the company putting its new AI *everywhere*, but mostly because viral user-generated AI images are now ubiquitous on its apps. My latest for @FastCompany: https://www.fastcompany.com/ ... Ali Alkhatib / @_alialkhatib : the instagram search page is like half advertisements for garbage AI stuff facebook is pushing, and that's before you even type anything in lmao.
Fast CompanyScott Nover
Context & Ripple Effects
Meta had already detailed how AI systems shape distribution across Feed, Stories and Reels in its recommendation-system disclosures, making the placement of AI material a core product-design issue rather than a standalone feature launch.
This report captures the user-side cost of combining ubiquitous assistant features with cheaply produced synthetic media. Later coverage of Meta removing viral AI profiles after backlash suggests that the company has had to weigh AI engagement experiments against perceived authenticity and usability.
First-order effects
Facebook and Instagram users encounter more Meta AI surfaces and AI-generated imagery in discovery flows, raising the immediate risk that search and feeds feel less relevant or harder to navigate.
Meta gains prominent in-app distribution for its AI tools, but the same placement can create dissatisfaction among users who see the added material as intrusive.
Second-order effects
Creators and advertisers must compete with a larger volume of synthetic posts and AI-focused placements for attention, making organic discovery less predictable.
If complaints persist, Meta faces pressure to tune recommendations, ad placement, labeling, or removal policies; the later retreat from AI profile experiments shows such AI-native formats can carry reputational costs.
Third-order effects
The episode points to a synthetic-supply problem: generative tools can increase the volume of publishable media faster than platforms can preserve relevance and social trust.
If platform-wide AI distribution becomes the default, differentiation may shift from simply generating content to ranking, provenance, and controls that let users limit unwanted AI material.
The trend: Major social platforms are using their distribution advantage to normalize generative AI, while confronting the quality and trust costs of abundant synthetic content.
The over-investment in machine learning initiatives, particularly LeCun's research group and building out and maintaining PyTorch, is actively degrading user experiences. Meta's actual ML needs are quotidian algos deployed at scale, not these “frontier” models. This push is fol…
Agreed, but Meta doesn't have product lines like Google where building out TF and bringing on the Keras team makes sense. Meta's ML needs are pretty banal, too, there's no need to be “cutting edge” in this way. …
Having worked there I don't agree that an alternate resource allocation would improve the apps. They are already fully staffed and scaling up headcount adds managerial overhead, not software quality. …
Things that make me lose my shit: Seeing a fellow language teacher on Facebook ask what AI site/app other teachers are using to “GENERATE THEIR CLASSES AND LESSON PLANS FOR THEM”. How about using your brain, expertise, your actual brain, your training and experience and YOUR BRA…
Grandma's Birthday Cake posts have gotten millions of engagements since it became relatively effortless to generate images like this. Share our post on this with your older relatives to explain it: https://theconversation.com/ ...
There's nothing that it does that has anything to do with things I've searched for on Instagram. I've never opened search hoping it would draw a picture for me
Meta: “How can we make IG better for users?” Users: “Chronological feed, show our posts to our followers, stop making us change workflows so you can compete with other apps...” Meta: “What if we just jam a shitty ChatGPT into the search bar?”
The Meta ecosystem is now flooded with AI—partly because is the company putting its new AI *everywhere*, but mostly because viral user-generated AI images are now ubiquitous on its apps. My latest for @FastCompany: https://www.fastcompany.com/ ...