Meta’s New Reels Algorithm Prioritises “True Interest” Over Engagement

Meta has fundamentally changed how Facebook Reels recommendations work, moving beyond watch time and likes to directly survey users about content relevance. The shift, announced January 14, 2026, has significant implications for affiliate marketers who have built strategies around engagement optimisation rather than genuine audience interest alignment.

The company's engineering team revealed that traditional recommendation signals achieved just 48.3% precision in identifying what users actually want to see. The new User True Interest Survey (UTIS) model improved that figure to 63.2% by asking users directly whether content matches their interests, then training algorithms on those responses.

For affiliates producing content on Meta platforms, the message is clear: gaming engagement metrics no longer delivers the distribution it once did. The algorithm now attempts to understand whether content genuinely serves viewer interests, with measurable consequences for both quality content creators and those producing generic viral material.

How the System Actually Works

Meta deployed randomised in-feed surveys asking users to rate content on a five-point scale responding to the question “How well does this video match your interests?” The company collects thousands of these responses daily, weighting them to correct for sampling bias and building what they describe as a comprehensive dataset reflecting real preferences rather than implicit engagement signals.

The technical implementation runs parallel to existing recommendation systems rather than replacing them. UTIS provides probability scores that videos will satisfy user interests, with those predictions influencing both late-stage ranking and early-stage retrieval.

What makes this significant is the scope of factors the model now considers. According to Meta's documentation, interest matching extends beyond topic alignment to include audio quality, production style, emotional tone, and viewer motivations. These elements prove difficult to infer from binary actions like clicks or shares, which is precisely why Meta built the survey-based approach.

The results from A/B testing involving more than 10 million users are substantial. High survey ratings increased 5.4% while low ratings decreased 6.84%. Total user engagement rose 5.2%, and content policy violations decreased 0.34%.

Why This Matters for Affiliate Content Strategy

The implications for affiliate marketers producing short-form video content are significant. Content optimised purely for watch time or likes may now receive reduced distribution if it doesn't genuinely match user interests.

Meta's announcement explicitly states the system increases delivery of high-quality niche content while reducing distribution of low-quality generic popularity-based recommendations. That shift directly affects affiliates who have relied on broad-appeal viral content rather than content tailored to specific audience interests.

The parallels to Google's crackdowns on low-quality affiliate content are striking. Both platforms have moved toward systems that attempt to identify genuine user value rather than content engineered to game engagement signals. The “spray and pray” approach of producing high volumes of generic content is becoming systematically disadvantaged across major platforms.

For affiliates focused on specific niches, this represents an opportunity. Content resonating deeply with smaller audiences on factors like audio style, production quality, and emotional alignment may now receive distribution advantages over content that achieves surface-level engagement from broader audiences.

Production Quality Now Directly Affects Distribution

The UTIS model's consideration of audio quality and production techniques marks a shift in what constitutes competitive content. Video content strategies for affiliates now need to account for production elements that weren't previously factored into algorithmic distribution.

This doesn't necessarily mean affiliates need professional studio setups. The model evaluates whether production quality matches user expectations and interests, not whether it meets some objective standard. Authentic, well-produced content that serves a specific audience may outperform polished generic content that doesn't genuinely address viewer interests.

The emphasis on mood and emotional tone also suggests that affiliates should consider how their content makes viewers feel, not just what information it conveys. Product demonstrations that create genuine enthusiasm, tutorials that reduce viewer anxiety about complex topics, or reviews that honestly address both strengths and weaknesses may perform better than content designed purely to maximise watch time through artificial tension or clickbait hooks.

The Longer Format Advantage

Performance data released alongside the UTIS announcement shows longer Reels over 90 seconds generating more than double the median views compared to TikTok videos. This suggests that when content genuinely serves user interests, viewers are willing to engage for extended periods.

The shift toward longer video formats across social platforms creates opportunity for affiliates who can create compelling extended content. Product reviews, detailed tutorials, and comprehensive comparisons become more viable when algorithms reward genuine interest satisfaction rather than punishing content that asks for more viewer time.

Facebook Reels now gets more than three times the median views of other video content on the platform, according to research from Emplifi cited in Meta's documentation. Combined with the UTIS improvements, this positions Reels as a primary distribution channel for affiliates willing to invest in content quality.

The Competitive Context

Meta's UTIS development occurs within intense competition in short-form video. Research from November 2025 found YouTube Shorts leading consumption at 56%, with TikTok and Facebook at 50% each. Instagram Reels captured 41% of usage among survey respondents.

The company has invested approximately $4 billion annually in Reels development specifically to counter TikTok, court records revealed. Reels reached a $50 billion annual run rate by the third quarter of 2025, with video time spent on Instagram increasing more than 30% year over year.

For affiliates, this investment translates to continued platform development and feature expansion. Meta's documentation indicates exploration of advanced modelling techniques including large language models and more granular user representations. The company also began using AI chat data for ad targeting in December 2025, affecting over 1 billion monthly users.

What Affiliates Should Do Now

The UTIS implementation suggests several strategic adjustments for affiliates producing content on Meta platforms.

Focus on genuine niche expertise rather than broad appeal. The system's emphasis on interest matching means content serving specific audience needs may outperform generic viral content designed for maximum reach. Affiliates who have built authority in particular product categories or interest areas should find their distribution improving relative to generalist competitors.

Invest in production quality that serves your audience. Audio quality, production style, and emotional tone now factor into algorithmic decisions. This doesn't require professional equipment, but it does require attention to whether production choices serve viewer interests or simply meet minimum standards.

Create content worth the time investment. With longer formats performing well and the algorithm attempting to identify genuine interest satisfaction, content that provides real value to viewers should receive distribution advantages. Detailed product reviews, comprehensive tutorials, and honest comparisons align with what the UTIS model attempts to surface.

Test and measure genuine engagement. Traditional metrics like watch time and likes remain relevant but are no longer the sole determinants of distribution. Monitor whether content generates follows, shares, and return viewers, signals that indicate genuine interest satisfaction beyond passive consumption.

The trends reshaping affiliate marketing consistently point toward quality and authenticity as competitive advantages. Meta's UTIS implementation formalises this shift into algorithmic infrastructure, making it systematically more difficult to succeed through engagement optimisation alone.

The Broader Signal

Meta's move to survey-based recommendation represents a broader acknowledgment that engagement metrics don't fully capture content value. The company explicitly states that traditional signals can be noisy and may not fully capture the nuances of what people actually care about or want to see.

For affiliate marketers, this validates strategies built around genuine audience service rather than engagement hacking. The platforms are getting better at distinguishing between content that satisfies users and content that merely captures their attention.

Content that demonstrates experience with products, provides genuine expertise, and addresses real audience needs aligns with what Meta's algorithm now attempts to surface. User-generated content approaches that feel authentic and relatable may outperform polished promotional content that lacks genuine interest alignment.

The affiliates who will benefit most from these changes are those who have already been producing quality content for specific audiences. For those who have relied on volume, broad targeting, and engagement optimisation, the UTIS model represents a significant challenge to existing strategies. The algorithm is now actively trying to identify and demote content that captures attention without serving genuine interests.

Meta's announcement doesn't eliminate the need for understanding platform mechanics and optimising content for distribution. But it does shift the optimisation target from engagement signals to genuine interest satisfaction, a change that favours affiliates who create content worth watching over those who create content designed to be watched.