Skip to content
TrustList
Blog

Will social media survive the AI change? What the evidence says

Editorial

By TrustList Editorial

Our overnight sample had no social referrers, while Meta's crawlers made 5.9% of page requests. Usage, labelling rules and the EU AI Act show what changes on social platforms, what survives and where to spend effort.

About Will social media survive the AI change? What the evidence says

Will social media survive the AI change? What the evidence says

Overnight from 24 to 25 September 2026 we counted every page request our edge proxy served for trustlist.uk over 9.3 hours: 77,221 in all. Of the 35,401 page views from browser-like user agents, not one arrived with a referrer from a social network. Three quarters (76.1%) carried no referrer at all, 23.8% came from our own pages, and 27 came from a Google search page.

In the same window, Meta's crawlers requested 4,519 pages, 5.9% of everything we served. That is twice as many requests as Bingbot made (2,180) and about twenty times as many as Googlebot (221). So in our sample the company that runs Facebook, Instagram, WhatsApp and Threads sent us no visitors we could see, and read our site more often than any search engine did. It is one overnight window, a sample and not a month, and much of our browser traffic is an automated browser farm rather than people. We allow AI crawlers in deliberately. But the direction is clear enough to ask the question properly: if social platforms increasingly read the web for their AI rather than send people to it, and their feeds fill with machine-made content, does social media as businesses have known it survive?

Our answer, from the evidence below, is that social media is not shrinking and is very unlikely to disappear. What is changing is what it does for a business. It is becoming less a way to send people to your website and more a place where people meet named individuals, ask each other for advice and consume video, with AI running through all of it. That calls for a different kind of effort, not no effort.

What Meta's crawlers are for

Meta documents its crawlers on its developer site, and the list explains our log. FacebookExternalHit fetches a page when someone shares a link on one of Meta's apps, to build the preview card. Meta-ExternalAgent crawls the web for uses such as training foundation AI models or improving products by indexing content. Meta-WebIndexer, in Meta's words, navigates the web to improve Meta AI search result quality. Meta-ExternalAds crawls for advertising and other business products. Meta-ExternalFetcher fetches individual links at a user's request, and Meta says it may bypass robots.txt for those user-initiated requests.

We have not split our 4,519 Meta requests by agent for this article, so we do not claim that all of them were for AI. But only one of those five agents exists to show a link to a person on a social feed. The others exist to train models, to answer questions inside Meta AI, or to target advertising. When a social platform reads your site, it is now at least as likely to be learning from it as to be sharing it.

Meta's own results show the scale of the bet. In its second-quarter 2026 results, published on 29 July 2026, Meta put its expected 2026 capital expenditure, including principal payments on finance leases, at $130 billion to $145 billion. Mark Zuckerberg said AI is accelerating the core business, powering the next generation of products and opening the door to new enterprise opportunities. A company spending at that level on AI infrastructure is not planning for its feeds to stay as they were.

The usage figures: social media is growing, not dying

Every primary source we opened points the same way. People are not leaving.

Meta's own count. Meta reported 3.60 billion family daily active people on average in June 2026, up 3% on a year earlier. Ad impressions across its apps rose 14% year on year and the average price per ad rose 12%. Revenue was $60.80 billion for the quarter, up 28%. Those are the figures of a business with more use and more advertising demand, not less.

US adults. Pew Research Center's social media fact sheet, published on 20 November 2025 from a survey of 5,022 US adults between February and June 2025, found that 84% use YouTube, 71% Facebook, 50% Instagram, 32% TikTok and 25% LinkedIn. YouTube use has risen from 73% in 2019 and Instagram from 35% in 2018. TikTok has risen from 21% in 2021.

News habits across 48 markets. The Reuters Institute's Digital News Report 2026, published on 16 June 2026, found that social media and video networks are for the first time the single most widely used way of reaching online news, used by 54% of people averaged across its markets. Facebook remains the biggest platform for news at 43%, reversing recent declines. Instagram is at 26% and TikTok, growing fastest from a small base, at 20%.

Communities. Reddit reported 130.3 million daily active uniques in the second quarter of 2026, up 18% on a year earlier, and 514.6 million weekly active uniques, up 24%. Its revenue rose 61% to $805 million.

Video and audio. Edison Research's Infinite Dial 2026 found that 78% of Americans aged 12 and over use YouTube weekly, and that 57% have both listened to and watched a podcast.

Whatever else AI is doing, it has not emptied the platforms. If anything, AI tools are arriving inside them. The more interesting question is what people will be looking at when they get there, and whether they will believe it.

What changes: feeds full of synthetic content

The first real change is the content itself. It now costs almost nothing to produce an image, a voice or a video that looks real. The platforms have responded with labelling rules, and each one has published its own.

YouTube requires creators to disclose when they have used AI to meaningfully alter or generate content that looks realistic: making a real person appear to say or do something they did not, altering footage of real events or places, or generating a realistic scene that never happened. Content that is clearly unrealistic, and minor aesthetic edits, are exempt. YouTube may show a label in the player, and says that creators who consistently fail to disclose may have a label applied for them, have content removed, or be suspended from the YouTube Partner Program.

Meta announced in February 2024 that it would label AI-generated images on Facebook, Instagram and Threads, detecting the industry signals in the C2PA and IPTC standards that tools from companies including Google, OpenAI, Microsoft, Adobe, Midjourney and Shutterstock embed. For photorealistic video and realistic-sounding audio, Meta requires the person posting to disclose it, and says it may apply penalties if they do not, and more prominent labels where content risks deceiving the public on important matters.

TikTok said on 10 March 2026 that its labelling efforts had helped label more than 1.3 billion videos. It combines creator labels, detection models, C2PA Content Credentials and its own invisible watermarks. It is also testing a control inside Manage Topics that lets people adjust how much AI-generated content appears in their For You feed, in the same way they can adjust dance, sport or food.

That last detail says a great deal. A platform that lets users turn down AI content has noticed that some users want less of it.

The law is catching up in Europe. The European Commission's AI Act page, updated on 3 August 2026, says the Act's transparency rules come into effect in August 2026, and that certain AI-generated content must be clearly and visibly labelled, namely deep fakes and text published to inform the public on matters of public interest. The Commission has published a voluntary Code of Practice on marking and labelling AI-generated content, and a set of icons that creators and publishers may use. The amending "digital omnibus" regulation, in force since 27 July 2026, delayed the high-risk rules but, according to the law firm Goodwin's reading, left Article 50's transparency obligations in place from 2 August 2026, with a grace period to 2 December 2026 for the machine-readable marking duty on generative systems already on the market. Goodwin notes fines of up to €15 million or 3% of worldwide turnover. A UK business that posts to audiences in the EU should assume the labelling expectations apply to its deepfake-style content there.

For a business, the consequence is simple. Undisclosed synthetic content is now a compliance risk on the platforms and, in the EU, a legal one. Disclosed synthetic content is allowed, but it is easy to ignore.

What changes: trust moves to people you can name

The second change is trust. The Reuters Institute found trust in news overall at 37% across its markets, trust in news on social media at 22%, and trust in answers from AI chatbots at 20%. It found 62% of people concerned about what is real and fake online, up four points. It also found that 46% of people get news from some kind of creator, and 27% from creators focused on news.

Read together, those figures describe people who use social platforms heavily and trust what they see there only a little. When trust in the feed is low, people fall back on sources they can identify: a person they know of, a publication with a name, a community where others vouch for each other. Reddit's chief executive, Steve Huffman, put it in his company's results release: in an increasingly automated web, the value of real human perspective has never been higher. That is a commercial claim, but Reddit's growth figures are consistent with it.

A third change we expect, but cannot measure well, is that more of the useful conversation moves into private spaces: group chats, invitation-only communities, direct messages. We have not found a primary source with a recent figure we are willing to print, so we state it as a judgement, not a finding. Our own log offers a small hint. Three quarters of browser visits arrived with no referrer at all, and links opened from messaging apps and email often arrive that way. Private sharing is real traffic that no analytics report will credit to "social".

What survives, and what probably does not

Put together, the evidence supports a fairly specific prediction.

What survives and grows:

  • Video, especially YouTube, which is also a search engine and a podcast platform.
  • Named people. Founders, engineers and specialists who post under their own names, with a history people can check. Their posts carry trust that a brand account and an AI-generated post do not.
  • Communities where people ask each other for advice, such as Reddit and specialist forums. These are read by buyers and, increasingly, by AI systems that summarise them.
  • Private groups and messaging, where recommendations travel between people who know each other.
  • Social platforms as advertising businesses. Meta's ad impressions and prices both rose in the last quarter.

What weakens:

  • Organic reach for brand pages that post generic content. When feeds fill with cheap synthetic material, a company page posting the same kind of material has nothing to stand out with.
  • Social as a traffic source to your website. Platforms have every reason to keep people in the app, and now have AI assistants that can answer inside it. Our sample recorded zero social referrers. We would not generalise from one site's overnight log, but we would not plan a business on social referral traffic either.
  • Undisclosed AI content. Labelling rules, detection and the EU's transparency duties make it a liability.

So social media survives the AI change. The version that fades is the one in which a business posts to a company page and expects visits to follow.

Deciding where to spend your social effort

This method is for a small B2B business, a software firm or an IT services provider, with a few hours a week. It works without any tool.

Step 1: find out where your buyers already talk

Ask your last ten customers three questions: where they first heard of you, where they go to ask peers about suppliers, and which people they follow for advice in your field. Write down the platforms and the named people. That list, not a general statistic, decides where you spend time. For most B2B firms it will point to LinkedIn, YouTube, one or two communities, and a handful of named individuals.

Step 2: choose one platform to own and one to listen on

Owning means posting regularly under a named person's account, not only a company page. Listening means reading and answering in a community where buyers ask questions, without advertising. Two is enough for a small team. Adding a third usually means doing all three badly.

Step 3: post evidence, not filler

The content that holds attention in a feed full of synthetic material is the content an AI tool cannot make: something you counted, a case you worked on, a mistake you made and fixed, a photo of your actual team or product. If you would not put your name to a post, do not publish it. If you use AI to help write or produce something, follow the platform's disclosure rules, and label realistic synthetic images, audio or video every time.

Step 4: measure the right things

Stop judging social effort by referral clicks alone. Track:

  • enquiries in which the buyer mentions a person or a post by name;
  • direct and referrer-less visits in the days after a post;
  • replies and messages from people in your target market;
  • how often your firm is mentioned, accurately, in communities you do not control.

Step 5: protect your own channel

Every platform above can change its rules, its reach or its AI features without asking you. Keep one channel you control: an email list, built by offering something worth subscribing to. Move the relationships you make on social platforms there when people are willing.

Step 6: decide what your site lets platforms read

Check which of Meta's, OpenAI's and other crawlers your site allows. Allowing a platform's AI crawler may mean your content helps answer questions inside that platform. Blocking it keeps your content out of its models and its answers. There is no universally right answer. We allow them, because we would rather be cited than absent, but it should be a decision, not a default someone else set.

A short checklist

  • One owned platform, one listening community, one email list.
  • Posts from named people, with evidence nobody else has.
  • AI-made realistic media labelled every time.
  • Success measured by named enquiries, not referral clicks.
  • Crawler access decided on purpose.

For how the rest of the attention picture has shifted, including answer engines and marketplaces, read where public attention comes from when search sends fewer clicks. For what has happened to search itself, see is SEO dying or changing.

Sources

Keep exploring

More on TrustList

Everything here links back to the same verified catalogue. Pick your next stop.