Off-Algorithm · 2026Chat groups are replacing open posts
Recommendations that used to circulate in public posts are increasingly happening inside closed chat groups, among people who already know each other or refer one another directly.
01 / READING
Choice starts to reappear
Off-Algorithm is the term Fresta uses for a movement already in circulation: people trying to reclaim part of the choice in environments organized by automatic recommendation.
Algorithms keep suggesting what to listen to, watch, read, buy or visit. They still offer convenience and scale. What begins to change is the willingness to let them decide alone.
That shift shows up when someone reviews and corrects the profile a platform has built about them; when a recommendation gains a recognizable author; when the shop, the fair, the community and local programming work again as places of discovery; when the absence of an automatic sequence starts to be presented as a value.
A handwritten card on a bookshop shelf makes three things visible: who chose, based on which repertoire, and where that recommendation can be questioned. A bestseller ranking shows aggregate popularity, but it neither states authorship nor allows individual correction.
Fresta's thesis: people still want convenience. What begins to lose value is personalization that shows no criterion and accepts no correction.
02 / WHAT CHANGED
The signal doesn't depend on a single scene
Part of discovery is happening outside the feed.
People are using proximity, trusted networks, physical presence and direct referral to find what automatic recommendation doesn't deliver. Each piece of evidence below shows one part of the movement. None, on its own, proves a widespread rejection of the algorithm. Together, they let us observe a new expectation: recommendation has to allow intervention, authorship or context.
Address and schedule travel by referral, not by search
Some events and spaces keep little public information. Addresses, schedules and invitations circulate after a referral or confirmation, not through open search.
Human recommendation scales inside the feed
More than 50 million books tied to the BookTok community were sold in 2025 across six European markets. The case combines algorithmic distribution with identifiable voices: authorship and community organizing discovery at scale, without leaving the platform.
Algorithmic awareness doesn't automatically produce resistance
A 2025 study in Frontiers in Psychology (Yuan, Shi, Su and Zhang) found a counterintuitive result: in an experimental information-navigation task with a defined goal, greater awareness of the algorithm increased compliance with it, not resistance. People may contest, negotiate or accept the system, depending on context.
Product · Instagram
Even the platform agreed to negotiate
Instagram adapted to that demand. In "Your Algorithm", people can now review and edit the topics used for recommendations, live across Reels, Explore and the main feed. It's the same demand for control and correction seen in the evidence above, now built into a platform that used to just suggest.
Product interface · Instagram · 2026Discovery without a feed trades scale for proximity and trust. Discovering outside the algorithm requires a social network and presence. For people without that access, automatic recommendation remains the main infrastructure available.
Implication for brands: brands can support communities, spaces and curators without immediately turning those relationships into one more tracking infrastructure.
03 / MECHANISM
In 1973, Stuart Hall described three ways of responding to a message: accepting the dominant meaning, negotiating it, or opposing it. He was talking about television. Today, much of what reaches people passes through recommendation systems, and those same three positions are useful again for reading what is happening.
Stuart Hall, Encoding and Decoding in the Television Discourse, 1973The ambivalence in recent research comes close to this reading: people don't respond in a single way to the systems that organize what they see. Some refuse outright, migrating to decentralized networks like Mastodon and Bluesky, to newsletters, closed groups and clubs outside the platforms, where the order of what appears isn't calculated. But the broader movement Fresta observes is the step before refusal: the shift from automatic acceptance to negotiation, when continuing to use the system coexists with the demand for more control over it.
04 / TENSION
The greater the dependence, the bigger the question about trust
The Reuters Institute's Digital News Report 2026 records a decisive contradiction.
For the first time, social networks and video platforms became the most-used way to access news globally, for 54% of people, ahead of television (52%) and news outlets' own sites and apps (51%). At the same time, only 22% say they trust the news they consume on social media, against 37% trust in news in general, the lowest global level since 2015.
People stay in these environments because they concentrate information, entertainment and convenience. The figure points to a broader tension, without covering every form of automatic recommendation: dependence on the infrastructure and trust in its criteria don't necessarily advance together.
Off-Algorithm gains relevance in that gap, when continuing to use the system starts to coexist with the demand for more participation in the choice.
05 / CONTEXTS
The impulse appears in markets that don't share the same cause
A phone-free event and an independent bookshop don't grow for the same reason. Even so, all of them create situations where discovery, selection and presence become less automatic.
What the data allows us to say: there is economic scale and institutional supply around experiences that reduce the automatic sequence or value local mediation.
What remains open: price, sustainability, nostalgia, sociability and commercial differentiation also explain these movements. The signal is in the convergence; the cause is not a single one.
+567%
Phone-free
Global growth of events advertised as phone-free between 2024 and 2025 on Eventbrite, with attendance growing 121%.
Eventbrite Newsroom, Offline by Design · April 2026
151%
Independent bookshops, over six years
Growth in American Booksellers Association membership, in the United States, up 19% in 2025 alone and 605 new operations in the year, across physical store, pop-up, mobile store and online.
American Booksellers Association, Annual Report 202506 / FIELD
Other ways of discovering
Fresta field notes from Berlin, in the first half of 2026, and a documented case in São Paulo.
Berlin · Bookshop and vinyl fair
Whoever chooses has a name, and the choice has an end
In an independent bookshop or at a vinyl fair, whoever decides what comes in is a person, not a ranking. The books on the shelf and the records in the crates are there because someone, responsible for that space, chose to bring them. Buying in a place like this begins with an act of trust: before flipping through anything, you're already trusting the curation of whoever assembled the selection. And that selection is finite, the opposite of the most-played list or the endless catalog of streaming. On Spotify, the choice dissolves into an infinite offer ranked by popularity; at the fair, it comes cut by someone's eye. It's that combination, curation with a name and a collection with a limit, that changes the experience of discovering.
Field · human, finite curation
Berlin · Street
The discourse already has a poster
Posters from the Radical Anti-Smartphone Front cover lampposts in Berlin, with their own name, manifesto and aesthetic. Born in Germany as an artistic and activist initiative, the collective treats the smartphone as a “poison” of social life and uses slogans like “love instead of liking” to provoke. From wheatpaste to sticker, it has already appeared in Paris, London and Brussels. It's the critique of the big platforms' automatic mediation gaining body, a name and public circulation across several European cities.
Field · discourse in circulation
São Paulo · Formosa Hi-Fi
You can tell where the choice comes from
Heirs to the Japanese jazz kissa, listening bars swap the streaming queue for a selection led live, in high fidelity. The format went from Japan to London, Milan and Sydney and reached São Paulo with Formosa Hi-Fi. Programming, acoustics and atmosphere give the music context back: who chose, by what criterion, and along which path the choice arrived there.
Case documented by Fresta · 202607 / INSIDE THE ARCHITECTURE
Why this signal, and why now
Fresta identifies changes in behavior before they become consensus and tracks each one until it grows, saturates or disappears. Off-Algorithm became a report because it grew faster and appeared on more fronts than most of the signals under observation in the same period.
The numbers below come from Fresta's architecture.
17 of 27
Most of what enters never becomes a signal.
Twenty-seven behaviors entered observation. Seventeen didn't sustain intensity and dropped out through extinction. Off-Algorithm is one of the ten still active. In a market where almost everything observed is announced as a trend, surviving Fresta's own extinction cut is already information.
1,134 articles
Measured recurrence, not a one-off reading.
Between June and July 2026, Fresta returned to this behavior more than sixty times and analyzed 1,134 articles, more than one reading a day since the signal was named. The criterion that confirms a signal is variety: research, press, industry and brand pointing in the same direction across more than one territory. Here, three: Brazil, the United States and Europe. Volume alone confirms nothing.
Three older signals, on the same axis
You can already tell where this tends to go.
The architecture links Off-Algorithm to three behaviors that have already traveled this path: Soft Refusal, now saturated, through the refusal of algorithmic demands; Interface Exhaustion, growing, through interface fatigue as a precondition; and Manual Residue, also growing, through the physical trace as an alternative to mediated discovery. Off-Algorithm is the youngest of a family whose eldest has already matured. It's its place in line, not a prediction.
Five other signals in Fresta's catalog share ground with this one: Hyperlocality, Local Friction, Permanent Ownership, Digital Folklore and Body as Territory. What brings them together is an editorial reading by the team, more than a link measured by the system. The stage of each is at frestasignals.com.
08 / WHAT CHANGES WHEN THE CRITERION CHANGES
Getting the suggestion right stops being the only measure of quality
When control and trust enter the evaluation of a recommendation, it also starts to matter whether the person understands why it appeared, and whether they can change what the system concluded about them.
Legibility begins to make up the product experience. Controls like “more of this”, “less of this” and “not for me” stop working merely as secondary settings and start to show that the choice remains open to intervention.
Identifiable authorship also gains relevance in curation contexts. It makes repertoire, responsibility and context visible, elements an aggregate ranking doesn't offer on its own.
For a brand, a platform and any selection-based business, the value of recommendation may start to include the clarity of the process: who chooses, which criteria organize the offer, and how much room there is for correction. Clicks and conversion measure immediate efficiency, but they don't show, on their own, whether someone broadened their repertoire, found something unexpected or came back out of trust in the curation.
Reach, in this scenario, becomes an insufficient measure of influence. An environment can remain efficient at distributing a message and, at the same time, fragile at building trust.
This debate is already under way. Regulators are beginning to require platforms to explain why something appears and to let people turn off automatic recommendation; researchers debate whether “opening the box” means simplifying the algorithm until it's transparent or, more realistically, keeping the system complex and opening only what affects each person: the criterion, the why, and the correction button. For brand and product, the takeaway is direct. Democratizing the algorithm is giving legibility and control where it matters, not revealing the formula. Whoever does it first turns control into trust, and trust into a relationship that lasts longer than a click.
Algorithms will keep organizing an important part of culture. The question that opens up is how much control people will come to expect over the criterion that organizes what they see.
09 / WHAT TO WATCH NOW
Expectation that lasts or a differentiation pitch
Off-Algorithm is still a contested signal: it may take hold or dissolve in the coming months. What's at stake is whether control, authorship and context become something people start to expect from any recommendation, or stay just as a differentiation pitch from a few brands.
Product and platform
Recommendation controls moving out of limited testing and into everyday use; a chronological or history-free option presented as a central benefit, not as a hidden alternative mode; recommendation explanations more specific than generic formulas like “because you liked”.
Market and behavior
Signed curation gaining ground beyond media, fashion and entertainment; growth of direct relationships, own lists, clubs, communities, local programming; choice driven by transparency and control, not only by accuracy or convenience.
10 / METHOD
How Fresta arrived at this reading
Fresta doesn't treat every occurrence as a signal. A signal only enters a report when the same impulse reappears, with variation, across contexts and sources of different natures.
01
Listening
Continuous monitoring of platform, publishing, community and direct cultural observation.
02
Detection
Identifying repetition with variation, before the pattern becomes consensus.
03
Naming
Editorial reading of what scattered occurrences start to mean together.
04
Positioning
The signal is positioned as emerging, growing or saturating, according to circulation and diversity of context.
05
Extinction
Signals that stay below the intensity threshold for twelve consecutive cycles leave the report automatically, without editorial review.
This reading combines platform product, behavioral research, market data and field notes. The materials are not equivalent and serve different functions within the reading. A platform launch signals institutional recognition, but it also reflects corporate interest. Market data helps to size scale, within the limits of its own methodology. Research observes conduct in a bounded context. Field notes reveal the concrete form of the signal, without representing its reach on their own. The interpretation arises from the convergence of these layers.
Current limit: the evidence confirms growing tension around choice, trust and recommendation. It does not yet allow us to claim a mass abandonment of the platforms, nor a single cause for all the behaviors gathered here. The field is urban and selected, and the observed convergence doesn't rule out explanations particular to each market.
11 / SOURCES
Sources and limits
External evidence was verified against primary or institutional sources. Read with care: platform announcements show corporate recognition of the problem, still far from mass adoption; company and association metrics describe only the universe they themselves observe.
- Instagram, “Control Your Instagram Reels Algorithm”, released December 2025 and updated in 2026 for Feed, Reels and Explore.
- Spotify Newsroom, “A New Era of Personalization: Shape Your Taste Profile”, 13 Mar 2026.
- TikTok Newsroom, “#BookTok community helps sell more than 50 million books across Europe”, 19 Mar 2026. Data attributed to NielsenIQ BookData and Media Control.
- Yuan, Shi, Su and Zhang, “Resistance or compliance?”, Frontiers in Psychology, 2025.
- Stuart Hall, Encoding and Decoding in the Television Discourse, Centre for Contemporary Cultural Studies, University of Birmingham, 1973.
- Reuters Institute, Digital News Report 2026, 16 Jun 2026.
- Eventbrite Newsroom, “The Rise of Phone-Free Experiences in 2026's Analog Era”, analysis published April 2026. Platform-internal metric.
- American Booksellers Association, Annual Report 2025, published May 2026. Membership and new-store data.
- Fresta Signals. Field notes in Berlin and São Paulo, first half of 2026; own images and observation notes.
- Fresta Signals. Internal data from the signal-tracking system, June and July 2026. The numbers describe Fresta's corpus and internal rules, not population prevalence.
External sources checked on 27 July 2026.
Conclusion
Algorithms will keep suggesting. What's in dispute is how much of the choice becomes negotiable again.
Off-Algorithm records a still-early change in how recommendation is evaluated. The suggested content still matters, but the criterion that organizes that suggestion starts to enter the experience of trust itself.
If you forget the rest of this report, remember this: people are reclaiming room to take part in the choice within the recommendation systems they already use.
Correction control, recognizable authorship and context make the way a choice was built more visible. There's still no evidence of a widespread abandonment of the platforms. There are signs that convenience, on its own, may stop sustaining the relationship.
This is the stage where Fresta works: when occurrences still look scattered, but already begin to reveal a pattern capable of affecting product, experience and decision.
Fresta Signals
frestasignals.com · contact@frestasignals.com
Off-Algorithm
Download Report ↓