‘Digital-first’ is not Instagram-first

The question worth asking first

When a buyer asks an AI engine about your category, can the machine read anything you have made?

It is an awkward question for a brand that has spent years building a following, because the honest answer is often no. The work exists; it is simply in a form the engine cannot use. That gap, rather than effort or budget, is what tends to decide whether a brand turns up in the answer.

The gap has a mechanical cause, and it is worth being specific about it.

How an engine reads a platform

An AI engine builds an answer from text it can retrieve and then quote. Both steps matter.

Instagram is image-first and sits mostly behind a sign-in. The meaning of a post usually lives in the picture, and the picture lives on a page a model often cannot reach. Where a caption carries the words, some of that meaning becomes readable; where it does not, the post is effectively dark to the engine.

The platform has opened a little. From 10 July 2025, public content from professional accounts began appearing in Google’s results — but only public, professional posts, only from 2020 onward, and only where the words sit in a caption rather than the image. The opening is real and also narrow. A feed can have considerable human reach and still offer an engine very little to work with.

What the engines actually cite

The surfaces that come up most across the major studies tend to be the open, text-first ones.

Semrush’s analysis ranks Reddit the single most-cited domain across ChatGPT, Google’s AI Mode, Gemini, Perplexity and AI Overviews, with YouTube and LinkedIn close behind. A separate Meltwater study of 9.5 million citations places LinkedIn as the second most-cited source in AI answers, behind YouTube.

A detail from the Meltwater work is worth pausing on: a little over half of those LinkedIn citations came from members with fewer than ten thousand followers. The engine appears to reach for the clear, useful post rather than the popular account — which suggests the thing being rewarded is readable substance, not reach.

What these surfaces share is plain enough. Open access, real text, transcripts, named people, the occasional structured list. What sits low on every list is the image-first, login-walled feed.

Where AI tends to pull from

None of this is an argument against Instagram. It builds desire, proof and brand in a way little else does, and it sells. It is an argument about balance. A brand whose only digital muscle is a walled feed is strong precisely where an engine cannot see it.

Why the click no longer rescues the grid

The older defence — that a strong feed still earns the visit — has weakened along with the click itself.

A 2025 Pew study found people clicked a result about 8% of the time when an AI summary appeared, against 15% without one. By 2026, roughly two-thirds of Google searches ended with no click at all. The win is less often a profile visit now, and more often a mention inside the answer.

That shifts the burden. To be named, a brand has to have left something an engine can read and verify — and a feed of images, however good, leaves very little of it.

What “digital-first” was supposed to mean

The phrase was meant to signal modern and measurable, found where buyers look. It has narrowed.

Somewhere along the way, for a good number of brands, digital-first collapsed into “we post a lot on one app.” The label kept its confidence while the strategy lost its breadth. Being genuinely digital-first in 2026 tends to mean being readable and citable across the surfaces an engine consults before it answers — open text, video transcripts, community threads, an owned page — rather than concentrated in a single walled feed.

There is a wider pattern beneath this worth naming only briefly. Princeton’s GEO research found that content carrying citable statistics earned roughly 41% more AI citations than the same content without them, while simply adding words did nothing. Substance a machine can verify, in text it can read, is what tends to be rewarded. A grid of images supplies neither — which is the same problem seen from a different angle.

A more even footing

The correction is not to leave Instagram. It is to stop letting it stand in for the whole of digital.

Anything a brand wants cited — its point of view, its facts, the reason to choose it — needs a plain-text home on the open web, not only a caption inside an app. Experts can be made citable where engines already look, which on the evidence means a real, specific point of view on LinkedIn and honest, useful answers in the community threads of the category. Video earns its citations through transcripts, so the words are worth posting outside the player. And one clean, answer-first page gives the engine somewhere to confirm a claim with the brand’s name on it.

Done this way, digital-first becomes plural again: text and video and community and an owned page, arranged so that whichever surface an engine happens to read, it finds something. Holding those surfaces in some deliberate relation — the work we tend to call Narrative Architecture — is what keeps a brand from being loud in one place and silent everywhere the answer is written. Instagram is one instrument. The point is to play more than one.

Frequently asked questions