This post is for our Paid Subscribers. If you haven’t subscribed yet,
Clarifying Questions
Before diving in, I would want to align on scope, since “search” on a visual, feed-based platform like Instagram can mean text search, tag and topic search, or true visual and multimodal content search, and each of those is a fundamentally different build with very different cost and timeline implications.
Are we searching primarily for accounts and hashtags, or do we also need content-level search across captions, audio, and visual elements within photos and Reels? This determines whether we are extending the existing keyword-based search index or building a much heavier computer-vision and multimodal retrieval system from scratch, a difference of months versus years in engineering timelines.
Is this meant to compete with Google and TikTok as a discovery-and-answers surface (”best coffee shop near me”), or is it primarily meant to help users find people and content they already have some intent about? Instagram’s Explore surface and current search already lean toward the latter. Expanding into the former is a much larger strategic bet with a very different success bar.
What is the primary monetization angle: ad-supported search results, or purely an engagement and retention play with no direct near-term revenue? This affects how aggressively we can insert sponsored content into results early without triggering user backlash or trust erosion before organic quality is proven.
Should search results be personalized per user based on their social graph and interest signals, or should they trend toward a neutral, query-relevance ranked list? Instagram’s core value proposition is personalization, but an over-personalized search experience can feel like an echo chamber and undermine genuine discovery for new topics and accounts.
Are we designing for a global audience across all languages and scripts, or starting with English-first markets? Multimodal search, especially visual understanding and audio recognition, has very different accuracy and recall rates across languages and regions, so the scope decision here meaningfully changes what “done” looks like at launch.
For this answer I will assume we are building a unified search experience that spans accounts, hashtags, places, and content (captions, audio, and visual elements in photos and Reels), personalized using the user’s existing graph and interest signals, monetized eventually through sponsored results but not in the first release, and launched English-first with staged language expansion into other high-value markets.
Product Description
Instagram launched in October 2010 and was acquired by Facebook (now Meta) in April 2012 for approximately $1 billion, a figure that in retrospect stands as one of the most consequential acquisitions in technology history. As of Meta’s most recent public reporting, Instagram has over 2 billion monthly active users globally, making it the fourth-largest social platform in the world by active users. Instagram’s business model is almost entirely advertising, and analyst estimates consistently place Instagram’s annual ad revenue above $50 billion, accounting for roughly 30 to 35 percent of Meta’s total advertising revenue. Beyond core feed and Stories ads, Instagram has built a growing creator commerce layer including in-app shopping, affiliate tools, and paid subscriptions for creators, but advertising remains the overwhelming revenue driver.
Despite this scale, Instagram’s search product has not kept pace with the rest of the platform. The current search bar is dominated by exact-username and hashtag matching, with very limited ability to handle natural-language queries such as “cheap sushi in Tokyo” or “minimalist apartment decor ideas,” use cases that have visibly and publicly migrated to TikTok. Multiple industry surveys published between 2022 and 2024 report that over 40 percent of Gen Z users in the United States now use TikTok as a primary search engine for discovery and recommendations, effectively treating TikTok the way older cohorts treat Google. Instagram’s own product leadership, including former head of Instagram Adam Mosseri, has publicly acknowledged that search is one of the weakest and most under-invested surfaces in the app.
The strategic tension is significant. Instagram’s core content unit, an image or short video with a caption and hashtags, contains far less explicit, indexable text than TikTok’s spoken-word-heavy videos or a traditional webpage, making keyword search inherently harder. At the same time, the opportunity cost of ceding “search as discovery” to TikTok and Google is enormous given how much ad inventory is tied to session time and query intent. Meta does have substantial assets to draw on: world-class computer vision models used across Facebook and Instagram for content moderation, ad targeting, and accessibility (the AI-generated alt text on photos uses these same models), plus Meta AI, the company’s large language model layer. Building a real search engine, one that understands visual content, audio, and natural-language intent rather than just matching literal strings in captions and hashtags, is the single highest-leverage way to reclaim that discovery behavior inside Instagram itself.
Define Goal
The core problem is that Instagram’s current search only reliably answers “find this specific account or hashtag” queries, while users increasingly want to ask open-ended, intent-driven questions such as “minimalist apartment decor ideas” or “date night outfit for fall” that the underlying content actually contains the answer to, but the current text-matching infrastructure cannot surface.
My goal is to make Instagram search capable of answering open-ended, intent-driven queries as reliably as it already answers exact-match account and hashtag queries, so that users default to Instagram rather than TikTok or Google for visual discovery.
The north star metric I would choose is weekly successful searches per active searcher, defined as a search session in which the user clicks into at least one result and remains engaged with that result (a post, Reel, or account profile) for 10 or more seconds, measured within a rolling 7-day window. I prefer this metric over raw search volume or search DAU because volume alone rewards a broken search experience just as much as a great one. A user who searches five times and clicks nothing is counted identically to a user who searches five times and finds exactly what they needed. Requiring a genuine engaged click after the search directly measures whether the query was actually answered well, which is the real product outcome we are trying to drive and the one that will compound into advertising revenue and session-time growth over time.
User Segmentation
The Intent Searcher (18 to 34, discovery-mode browser): arrives on Instagram with a specific but open-ended need such as “nail art ideas for summer” or “best ramen spots in Austin,” currently gives up and switches to TikTok or Google when Instagram search returns thin or irrelevant results. This is the highest-value segment if captured because they bring exactly the kind of query intent that advertisers pay a premium for. Once trust in result quality is established, churn risk is low since switching apps mid-intent is friction the user would happily avoid.
The Social Searcher (all age bands, relationship-driven): uses search mainly to find specific people, brands, accounts, or events they already know about by name. This segment is reasonably well served by the current search experience and represents lower incremental upside. However, they must not be degraded by changes aimed at the other two segments, since any regression in exact-match retrieval quality would be quickly noticed and punished in app store reviews.
The Creator and Business Searcher (25 to 45, professional use): searches for competitor accounts, trending audio, top-performing hashtags, or emerging aesthetic categories to inform their own content strategy and posting calendar. This segment values accuracy and freshness of trend data above nearly everything else. The indirect value here is substantial because better creator tools and better trend intelligence drive more content supply onto the platform, which in turn makes search results richer for everyone else.
I would focus first on the Intent Searcher, because this segment represents the clearest unmet need, the most direct competitive threat from TikTok, and the largest addressable upside in session time and ad inventory. Critically, improvements built for open-ended intent queries, including better visual understanding, caption semantics, and natural-language parsing, will also incidentally improve results for the Social Searcher and Creator Searcher, making this the highest-leverage starting point across all three segments simultaneously.
Pain Points
Synthesizing public statements from Instagram leadership about search shortcomings, App Store review themes and ratings patterns, comparative product analysis against TikTok’s search surface, and Meta’s own transparency reporting on content quality challenges, five pain points stand out clearly.
