This month in Social Intelligence: Aug 2026
The social intelligence industry is changing faster than most people have time to follow. Companies merge, people move, platforms evolve, AI capabilities appear almost weekly, but not every announcement changes the discipline. Each month, we'll cut through the noise to identify the developments that genuinely matter. We'll explain what changed, why it matters, and what it tells us about where social intelligence is heading next.
August 2026
The internet is becoming a feedback loop
Social intelligence has always relied on a relatively simple assumption. That people do things online, leave traces of those behaviours behind, and practitioners analyse those traces to understand people, culture and society. But it's never been quite that simple, algorithms have always shaped what becomes visible, platforms determine what data can be accessed, people behave differently depending on the environments they inhabit, and now something more complicated is now happening.
AI systems are analysing digital evidence to answer questions and make recommendations. This has resulted in brands starting to think about how they can influence the sources those systems rely upon, AI-generated content is becoming part of the environment humans encounter and respond to and algorithms continue to determine which human expression gets amplified in the first place.
The result is an increasingly recursive information environment in which humans, platforms and machines aren't simply producing separate signals; they are influencing one another.
For social intelligence, that creates a much bigger challenge than whether AI can analyse social data, we need to decide what counts as evidence when humans, algorithms, and machines are all participating in its production?
When social data becomes evidence for machines
Reddit has become an increasingly interesting battleground for AI visibility. As organisations try to understand what influences answers produced by ChatGPT, Google and other AI systems, attention has shifted towards the third-party sources those systems retrieve and reference. Reddit conversations, reviews, forums, YouTube content and other forms of user-generated content have consequently gained a second audience - the machines.
This is starting to change how organisations think about participating in those environments. Recent reporting has examined brands attempting to influence Reddit conversations as part of their efforts to improve visibility within AI-generated answers. New technology providers are also emerging around this problem, including tools designed to identify the online conversations believed to influence AI recommendations and help organisations decide where to intervene. At the same time, the sources preferred by AI systems aren't stable. Analysis circulating this month suggests the prominence of Reddit within ChatGPT citations has changed considerably, with greater emphasis potentially being placed on structured and authoritative sources. And it is likely that the precise source mix will continue to change.
Why it matters
For years, social intelligence practitioners have treated online conversation as evidence to be observed and interpreted. But now we need to understand what happens when organisations deliberately change that conversation because they know machines are observing it too.
A Reddit thread could simultaneously be a genuine conversation between people, a source of social intelligence, an input into an AI-generated answer and a target for organisations attempting to influence that answer. That raises questions about authenticity, integrity and provenance that go considerably beyond GEO. Practitioners may increasingly need to understand not only what people are saying, but why particular pieces of digital evidence exist and who, or what, they were intended to influence.
Algorithms don't simply distribute culture
New research shared this month raises another challenge to how we interpret digital behaviour. An analysis of Twitter/X highlighted by NYU Professor Jay Van Bavel found a difference between content people actively choose to follow and content selected for them algorithmically. Posts from followed accounts were more closely aligned with users' stated values, while algorithmically recommended content showed greater misalignment.
It's tempting to think of social platforms as enormous collections of human expression. But the environment we observe isn't created by humans alone, recommendation systems decide what receives attention, what travels further and what people encounter next. So, the platform itself is participating in the production of the observable environment.
Why it matters
For social intelligence, this complicates the relationship between visibility and significance. Something appearing frequently in a feed doesn't necessarily mean people sought it out., high engagement doesn't necessarily mean endorsement, and widespread exposure doesn't necessarily tell us what people value.
Increasingly, we may need to distinguish between human expression, human behaviour and platform behaviour when interpreting social data. All three are not evidence of the same thing, although they can be meaningful.
What if people actually like AI slop?
A journalist recently experimented with using LinkedIn's AI writing capabilities to create deliberately generic AI-generated posts and reported a noticeable increase in engagement. While this is only one experiment and certainly doesn't prove that audiences universally prefer AI-generated content, it does start to raise some uncomfortable questions because much of the discussion around so-called "AI slop" assumes people don't want it.
What if sometimes they do? Or, perhaps more importantly, what if the distinction stops mattering behaviourally?
If AI-generated content attracts human attention, generates comments, gains followers and gets further amplified by recommendation systems, it becomes part of the social environment whether practitioners consider the original content authentic human expression or not.
Why it matters
The evidential distinction becomes important. An AI-generated post isn't evidence of human expression in the same way as something written by a person, but a person's decision to engage with it is human behaviour. And if that engagement causes an algorithm to amplify the content, we now have another signal again: platform behaviour. Treating all three as simply "social data" risks collapsing very different forms of evidence into one dataset.
The challenge for social intelligence may therefore become less about removing AI-generated content and more about understanding what kind of evidence each digital trace actually represents.
AI becomes part of the threat environment
Cyabra announced this month that two national security agencies in Europe had signed contracts with the company, worth more than $500,000 in combined annual revenue. The work focuses on foreign information manipulation, coordinated bot networks and GenAI-amplified narrative threats. It is another reminder that generative AI isn't only changing the tools used to analyse information environments, but it is also changing the environments themselves.
Why it matters
Social intelligence, information integrity, disinformation analysis and threat intelligence have long overlapped, and Generative AI may make those boundaries even harder to maintain. For example, synthetic content can now be produced at scale, coordinated networks can behave more convincingly, and human and machine activity can become increasingly difficult to separate. For practitioners, identifying what is being said may therefore become only the beginning of the analytical problem. Understanding who or what produced it, how it spread, why it became visible and whether it represents genuine human behaviour may becomes important depending on the questions you are trying to answer.
MCP becomes infrastructure
Last month, we wrote about the emergence of headless social intelligence as Meltwater, Hootsuite and others began connecting specialist technology directly with AI assistants through Model Context Protocols. A month later, across the technology ecosystem, MCP integrations continue to appear as platforms make their capabilities accessible to AI assistants and agentic workflows. The shift has become significant enough that SI Lab has now added MCP capabilities as a searchable filter within the SITech landscape and database (search here).
Why it matters
It is interesting on how quickly the underlying MCP architecture is becoming normal. The social intelligence stack is starting to become a network of capabilities that machines can call upon rather than a collection of destinations and logins for practitioners to visit.
If we go back to July's question, what happens when the interface disappears. August then introduces the next one: what happens when machines can move across the intelligence stack while simultaneously participating in the information environment we're trying to understand?
For years, social intelligence has often treated the post, mention, conversation or account as the thing to be analysed (the unit of analysis). But a digital trace isn't simply a record of what someone thought or did, it is the product of an environment. Increasingly, understanding what a trace means requires understanding what went into creating it: the human who produced it, the systems that shaped what they encountered, the algorithms that determined its visibility, the machines that may have generated or interpreted it, and the organisations attempting to influence those systems.
Digital evidence isn't truth waiting to be collected. We need to understand how it was produced before we can decide what it tells us, and as humans, platforms and machines increasingly influence one another, that production process is becoming a feedback loop.
The people shaping the profession
Insider 50 winner Libba Z. Peromsik has joined Pinterest as Social Analytics Lead. Her move brings recognised social intelligence expertise into one of the platforms shaping how people discover, plan, and express intent online.
Submit Industry News
Help us keep the community informed.
Have you launched something new, changed roles, published research, announced an event, or have company news relevant to the social intelligence community? We'd love to hear about it, send your details here.
Sources and further reading
When social data becomes evidence for machines
Katie Deighton, Brands Suddenly Care About Reddit. Redditors Don’t Return the Feeling, WSJ
FancyAI, Agentic Social Engagement →
Algorithms don't simply distribute culture
Ziv Epstein, et al, Value misalignments in X’s feed algorithm is a reflection of value tensions in engagement, Proceedings of the National Academy of Sciences (Open Access)
What if people actually like AI slop?
Katie Notopoulos, I stared deep into the heart of LinkedIn AI cringe slop. Darkness (and engagement) stared back, Business Insider
AI becomes part of the threat environment
Cyabra, LinkedIn