
Social platforms as search surfaces: what the evidence supports
People now look for products and answers on social platforms and in AI summaries as well as on Google. What the platforms have published about how content gets found there, what is still unproven, and what follows for the words in a title or caption.
In December 2025 Instagram cut the number of hashtags allowed on a post or Reel from 30 to 5. A long tag list had been the usual way to tell a platform what a post was about. With 5 tags, most of what describes a post now sits in the caption and in the content itself.
That change is one sign of how discovery on social platforms is moving closer to search.
Where people look
Ofcom's 2026 survey of UK adults found that 82% of 16 to 24s watch videos on social media, and that 83% of that age group use TikTok. It also found adults exploring fewer new websites. 14% said they use lots of new sites, down from 24% a year earlier, and 40% use none.
AI summaries are part of the same picture. 75% of online adults read AI search summaries at least sometimes, and 42% read them often or always.
The survey doesn't measure searching on social platforms. The closest public figure is older and from the US. In July 2022 a Google senior vice president, Prabhakar Raghavan, said the company's research found that almost 40% of young people looking for a place for lunch went to TikTok or Instagram instead of Google Maps or Search. Google said the research surveyed US users aged 18 to 24.
What the platforms have published
Several platforms have described retrieval that works on meaning. Spotify published a paper in March 2026 on podcast discovery, and LinkedIn published one in October 2025 on retrieval for its feed. Both describe matching the meaning of content to what a person is looking for.
Microsoft's AI Performance report in Bing Webmaster Tools shows site owners the retrieval phrases its AI used to find their pages. The post on reading that report covers it in detail.
TikTok's own description of its recommendation system is narrower. It lists what a viewer watches, likes and shares, along with details of the video such as its soundtrack and hashtags.
What isn't documented
Much of the advice in circulation goes further. It says platforms transcribe every spoken word and scan video frame by frame. TikTok's and Instagram's own pages on ranking, checked in November 2026, describe neither.
Another common claim is that a platform penalises an account for covering more than 1 topic. No platform source for that turned up either during planning for a podcast launch. These claims may be true. Until a platform confirms them, they are a weak basis for a content plan.
What follows for titles and captions
Fewer tags and meaning-based retrieval both put more weight on wording. The title and the caption are where a post says what it is about, and keyword research shows which words the audience already uses for the subject.
A podcast launch for a UK wellbeing brand shows this in practice. Each episode title leads with the topic in the wording parents search for, and the same titles go to YouTube. 2 episodes planned around a key search term drew about 4 times the median episode's plays, and Spotify impressions grew from about 11 a day to 35 a day over 2 months. Both guests are well known, which also moves the numbers on a show that size. The case study has the detail.
One audit across the surfaces
For a brand deciding whether social search or AI search needs a budget of its own, there is one audit. It covers how the brand shows up on the search surfaces its customers use, starting from what they search for. The work is mostly the work that was already on the list.
Sources
- Ofcom, Adults' Media Use and Attitudes report, published 2 April 2026
- Business Today, Instagram limits hashtags to 5 per Reel and post, 19 December 2025
- TechCrunch, Prabhakar Raghavan at Fortune Brainstorm Tech, 12 July 2022
- D'Amico, De Nadai, Chandar et al. (Spotify), "Deploying Semantic ID-based Generative Retrieval for Large-Scale Podcast Discovery at Spotify", 18 March 2026
- Ramanujam, Alonso, Kataria et al. (LinkedIn), "Large Scale Retrieval for the LinkedIn Feed using Causal Language Models", 16 October 2025
- TikTok, How TikTok recommends content, accessed November 2026
- Instagram for Creators, Algorithms and ranking, accessed November 2026


