Two days ago Google made Search Console reporting available worldwide for Instagram, TikTok, X and YouTube. For the first time you can see how content on a profile you do not own performs in Google Search · without owning a website at all. Which sharpens a question plenty of owners have been asking for years: do I actually need a site any more? We had 19 days of our own measurement sitting on a server, so we went and counted. Here is what AI engines really cite.
the date Google made Search Console “platform properties” available worldwide, letting anyone connect an Instagram, TikTok, X or YouTube account and see how that content performs in Google Search (Search Engine Journal, Search Engine Land) · the staged rollout began 7 July
of everything three AI engines cited, in Mode's own 19-day measurement, was an ordinary owned website · not a social profile, not a directory (Mode Marketing, 6,394 citations, July 2026)
the share of those citations that went to Instagram, TikTok or X combined · 14 citations out of 6,394 · the three visual-feed platforms Google just started reporting on
TikTok citation in the entire dataset, and it was an ads.tiktok.com marketing blog post, not a creator profile
YouTube's share · the only one of Google's four new platform properties that AI engines cite in any volume, and they cite specific videos rather than channels
YouTube's share when the question is about a trade or industry rather than a brand · in that slice Instagram was 0.35% and TikTok was zero (1,158 citations)
Mode's own measurement, method disclosed in full below so it can be checked. Aggregates only · no individual business is named or ranked.
The change itself is good and worth using. Search Console “platform properties” let you verify an Instagram, TikTok, X or YouTube account and see performance, insights and query-level data for how that content shows up in Google Search, Discover and News. Google started a staged rollout on 7 July 2026 and opened it to everyone on 29 July. It is free, and it reports on something most owners genuinely could not see before: what people were searching when they landed on your posts.
But notice exactly what it measures. It measures your social content's performance in Google Search. That is one surface. The surface where a growing share of buying questions now get answered · an AI assistant assembling a recommendation from a handful of sources · is a different one, and nothing about the Search Console change touches it. The two get conflated fast, because both have the word Search in them. They are not the same pipe.
So the useful question is not “can I now measure my Instagram”, which is settled and yes. It is: when an AI is deciding who to name, what kind of source does it reach for? That we can answer with our own numbers rather than an opinion.
Mode runs a citation harness: the same bank of 53 buyer questions, put to three AI engines on a schedule, with every source they cite recorded. Between 13 and 31 July 2026 that produced 6,394 citations across 1,280 distinct domains, once we set aside the questions about Mode itself (more on why in the method note). We sorted every cited domain into three buckets · an ordinary owned website, a social or video platform, or a directory such as a trade listing.
Owned websites took 82.6%. Social and video platforms took 15.6%. Directories took 1.8%. But the 15.6% is where it gets interesting, because it is not spread the way the Search Console announcement might lead you to expect. Of Google's four new platform properties, YouTube alone accounts for almost all of it at 5.52%. Instagram, TikTok and X together were 14 citations. Out of 6,394. That is 0.22%, and the single TikTok citation in the whole dataset was a marketing blog post on ads.tiktok.com rather than anyone's account.
The platforms that do get cited alongside YouTube are LinkedIn (5.08%) and Reddit (3.99%) · which is to say, the text ones. Places where a model can lift a sentence and attribute it. We did not test why, so we are not going to claim a mechanism. We will just report the shape: in this dataset the text-and-discussion platforms were cited roughly twenty times more often than the visual feeds.
We want to be fair to the social-first owner here, because the instinct is not stupid. If your customers are on Instagram and your work is visual, a strong feed earns real business, and it does so faster and cheaper than a website ever will. The new Search Console reporting makes that channel measurable for free. Use it. None of what follows argues against any of that.
What the numbers say is narrower and harder to dodge: a profile is not a source. When an AI assembles an answer about who to hire, it pulls from pages it can read, quote and attribute · and in our measurement that meant a website 82.6% of the time. Narrow it to the questions that look most like a real buyer asking about a trade or an industry, and 75.9% went to owned websites with another 6.0% to directories. Instagram in that slice was 0.35%. TikTok was zero. If a feed is your entire presence, you are not competing badly in that surface. You are absent from it.
The practical version is unglamorous. A small owned site whose pages answer the questions buyers actually ask, in text a machine can read without executing JavaScript, plus the directories your trade already uses, plus the social accounts you are already good at. YouTube is the one platform worth treating as a genuine content surface rather than a shopfront, on the evidence above · but note that engines cite videos that explain something, not channels that exist.
Confidence: moderate, and deliberately not stronger. The direction of this finding is stark enough that we do not think a bigger sample flips it. But our question bank is a UK marketing, agency-selection and small-business set · a query bank built around fashion, beauty or food would very likely surface more Instagram than ours does, and we would expect that. What would change our mind: a materially different question set showing visual-feed citations in double digits, or the engines adding image and video understanding to their retrieval in a way that starts pulling feeds into answers. We re-run this harness daily, so we will know if it moves, and we will say so here.
Mode operates a citation harness that asks a fixed bank of 53 buyer questions to three AI engines (OpenAI, Perplexity, Anthropic) on a schedule and records every source cited. This analysis covers 13 to 31 July 2026: 571 runs that returned citations, 8,316 citations in total across 1,464 distinct domains. Every cited URL was reduced to its host and classified into one of three buckets by a fixed list · social or video platform, directory or marketplace listing, or owned website (the residual). Headline figures exclude the six brand questions about Mode itself, which leaves 6,394 citations; those questions pull in profiles belonging to unrelated firms with similar names, which would inflate the social share and tell you nothing. That excluded slice is reported separately in the raw data. Four honest limits. First, this is our question bank, weighted to UK marketing, agency selection, cost, trust and trade verticals · it is not a general-purpose sample, and a differently-weighted bank would give different platform shares. Second, Perplexity cites on every run while the other two cite sparsely, so it contributes about 85% of the citation volume and the aggregate leans its behaviour; per-engine splits are in the raw data. Third, Gemini is in the harness but contributed zero citations in this window because its API quota is unfunded, so this is a three-engine result. Fourth, we measured what is cited, not why · the text-versus-visual reading is our interpretation of the pattern, not a tested mechanism. Google Search Console platform-properties dates, scope and metrics: Google Search Central, as reported by Search Engine Journal and Search Engine Land, 29 July 2026. Raw data and the classification lists: competitors/research/2026-07-31-social-vs-owned-citations.json. Corrections welcome at hello@modemarketing.ai.
Not if you want AI to recommend you. In Mode's July 2026 measurement of 6,394 citations across three AI engines, 82.6% of everything cited was an ordinary owned website. Instagram, TikTok and X together accounted for 14 citations · 0.22%. A social profile is genuinely useful for reach, proof and customer contact, and Google now reports on how it performs in Search. But when an AI assistant is assembling an answer about who to hire or what to buy, it overwhelmingly pulls from websites. If your only presence is a visual feed, you are close to invisible in the surface that is growing fastest.
They are a property type in Search Console that lets you connect a supported social or video account · Instagram, TikTok, X or YouTube · and see how that content performs in Google Search, Discover and News. You get performance, insights and achievements reporting, including query-level data. Google began a staged rollout on 7 July 2026 and made it available worldwide on 29 July 2026. It is a genuinely useful measurement tool. It reports on your social content's performance in Google Search · it does not make that content more likely to be cited by an AI assistant, which is a different surface entirely.
Partly, and it is the honest exception. YouTube was 5.5% of all citations in our measurement and 10.4% when the question was about a trade or industry, making it the only one of Google's four new platform properties that AI engines cite in any volume. But the pattern matters: engines cite specific videos that answer a question, not channels. A YouTube video that genuinely explains something can earn a citation. A channel that exists as a brand presence generally does not. Treat it as a content surface, not as a substitute for a site.
The likeliest explanation is that engines cite what they can read and quote. Reddit (4.0% of citations) and LinkedIn (5.1%) are text · discussions, posts and profiles that a language model can extract a claim from and attribute. Instagram and TikTok are primarily visual feeds whose substance lives in images, video and audio rather than indexable prose. We are reporting the measured pattern here rather than a mechanism, because we did not test causation: what we can say is that in 6,394 citations the text-and-discussion platforms appeared roughly twenty times more often than the visual-feed ones.
Get a small owned site that answers the questions buyers ask, in text, that a machine can read without running JavaScript. It does not need to be big. In the same dataset, when the question was about a trade or industry, 75.9% of citations went to owned websites and another 6.0% to directories such as trade listings · so the practical route in is your own pages plus the directories your trade already uses. Keep the social accounts and keep the new Search Console reporting, which is free and now tells you something you could not see before. Just do not mistake a profile you rent for a presence you own.
If your presence is mostly social, the first useful thing to know is what an AI assistant currently finds when someone asks about a business like yours. The AI Visibility Check answers exactly that: does AI mention you, can it read what you have, and the top three fixes · benchmarked against your three closest competitors.
AI VISIBILITY CHECK · £95Honest by design: we measure signals, not guarantees. How GEO works.