Empty Vessel in a Sea of Images

July 2026 7 min read

There are more images online today than any human could look at in a thousand lifetimes.

They flood every platform, every feed, every search result. Produced by cameras, studios, AI systems generating thousands per second. They are the primary way products are represented online, the face of services, the first impression of a brand, and they carry more commercial weight than almost any other element on a webpage. The volume is staggering and it is accelerating. What isn't accelerating is our ability to make sense of it, because almost none of these images carry any context with them.

Not because the context was never there. It was. When the web was built around text, an entire infrastructure emerged to make that text more understandable: URLs, meta tags, schema markup, structured data. So the context lives in the captions beneath images and in the descriptions on the page. Strip that away, remove the image from its written context, and what's left lacks intent, meaning, and value. An empty vessel, drifting in a sea of images exactly like it, with no memory of where it came from, who created it, or why it even exists.

Metaphorically speaking, it's like looking at an unannotated 100-year family photo, and not knowing who a single person is, where the photo was taken, or what their exact relationship to you is.

That gap - the absence of any persistent identity layer for images on the open web - is, in my view, one of the largest unaddressed problems on the internet. It was always there, but AI is about to make it impossible to ignore.

The Machine Sees a Shoe

Vision models can identify what's physically inside an image with remarkable accuracy. Show one a photograph and it returns objects, colors, scenes, compositions - instantly, confidently, at scale. This feels like understanding. It isn't.

Here's a hypothetical example.

A vision model sees an old boxing shoe. It returns: worn, shoe, leather, brown, laces, footwear.

Accurate. But almost entirely useless.

It has no idea the shoe belonged to Muhammad Ali. That it was worn at his final public appearance. That it's going to auction next month with a reserve price of $40,000. The machine sees what's in the frame. It cannot see what the image means, who made it, what it's for, or how it's permitted to be used. Those things don't live in pixels. They have to be declared.

Without that declaration, the machine guesses. It places the image in semantic buckets where it thinks it belongs, for an audience it thinks might care. It learns - slowly, over days - by watching who engages. Meanwhile the image sits in algorithmic limbo, reach stalls, and the window closes. The algorithm moves on. Not deleted. Permanently passed over.

“The algorithm moves on. Not deleted. Permanently passed over.”

This is the cold start problem. Every image on the internet begins with zero algorithmic understanding. And most of them never escape it, because the meaning that would have told the machine how to categorize it to the intended demographic was never attached to the image, or in time, became divorced from the context that briefly held it.

What Platforms Do With Your Images

Here's where large platforms only deepen the problem, and where the most common misconception lives.

Most people assume metadata is useless because it "just gets stripped." Stripped is either a misnomer or a strategic lie.

Every time you upload an image to a platform, it consumes everything traveling inside the file. The embedded metadata. The device fingerprint. The camera model. The GPS coordinates. The creation date. The editing software signature. They use it to build behavioral profiles, ad targeting vectors, trust scores, geographic patterns, and creator categories. They infer relationships, intent, commercial value, and audience fit - all from the data inside your file.

Then every field, every declaration, every piece of authorship and attribution you embedded disappears from the public-facing file. Locked inside their system.

“Your meaning becomes their data. Your intent becomes their inventory.”

Your image circulates the internet as an empty vessel, while the context that made it valuable lives in their walled garden, working for them.

It was never that machines couldn't read image metadata. It's that the platforms had every incentive to consume it privately and return nothing.

The issue doesn't stop there. AI has changed how discovery works entirely. Search used to be keywords typed into a box - a person browsing results, making their own choices. Now it's an agent. It crawls your site, builds a model of what you are and what you offer, and when someone asks a relevant question - in any form, in any language - it either surfaces you or it doesn't. Based on what it understood. There's no page 2. There's no "also consider." The AI made a judgment call, and you're either in the answer or you're not.

Who Feels This

This touches anyone who works closely with imagery, treating their website images as an extension of their products, services, and brand.

Photographers who watch their work circulate without credit. Designers whose output gets scraped and absorbed into systems that remember no one. Artists who spend weeks on a concept image, only to have it surface somewhere else as a derivative without acknowledgment. Small makers and product sellers who invest real time in product photography, only to find those images floating across the internet completely disconnected from their shop.

Take brands as a sharper example. A brand invests thousands - sometimes tens of thousands - in a single photoshoot. Professional photographers, location fees, talent, post-production, weeks of creative direction. Every image is viewed as an asset. And then production ends. The files get distributed to social teams, PR agencies, editorial partners, retail partners. And the imagery that was supposed to embody the brand becomes anonymous the moment it leaves the page.

We keep skipping the final step. Images should be treated as ambassadors, explicitly communicating themselves as extensions of the brand's products and services. Yet imagery goes out into the world as a floating object, with no memory of where it came from, who created it, or why it exists, working for no one but the platforms it passes through.

What VISID Does

VISID is an identity system for images - a set of tools built to establish and maintain discoverability, attribution, and provenance for your imagery. The approach was inspired by something much older than the internet: the instinct, practiced for thousands of years across cultures, to inscribe meaning directly onto an object so it survives the journey. Oracle bones. Clay tablets. Illuminated manuscripts. The artifact and its context as the same thing. VISID applies that instinct to images.

There are currently three core features.

VISID View

Before you stamp an image, you should understand what the machine already sees. VISID View lets you inspect any image - your own or one you've encountered online - and surface the metadata it's currently carrying. What's embedded. What's missing. What a platform's ingestion pipeline would actually read at the moment of upload. It's a diagnostic layer: see the gap before you close it.

VISID Stamp

Stamp is where declared identity begins.

You upload an image. VISID's AI reads it - understands it, not just labels it. But here's where it diverges from every vision model you've encountered. The AI doesn't work alone. You tell it what the image means, its significance and context.

Tell VISID that shoe belonged to Muhammad Ali. That it was worn at his final public appearance. That it's going to auction at $40,000. Now the AI understands the image in a completely different light, and everything it generates reflects that: the title, the description, the keywords, the structured data. The same pixels. Completely different meaning. Declared, not inferred.

That declared context gets embedded permanently - inside the file itself as metadata and echoed in the form of a JSON-LD for the webpage it lives on, a persistent public index, and drafted social captions - all tied to a fingerprint derived from the image itself.

The image now knows what it is. Who made it. What it's for. Whether AI was involved. Whether it can be used for AI training. And it carries that information wherever it goes.

When it hits a platform's ingestion pipeline, the answer is already there. No algorithmic limbo. No days of drift while the system figures out what it's looking at. It lands in front of the right audience from millisecond zero.

When an AI search agent crawls the page it lives on, three signals say the same thing - the structured data on the page, the metadata embedded inside the file, and VISID's public index. All aligned. All consistent. All pointing to the same declared truth. High confidence. Higher selection probability.

VISID Verify and Rehydration

Every image stamped through VISID does something that goes beyond the file itself. At the moment of stamping, the image is broken down - its structure, composition, color, tone, every visual relationship within the frame - and shaped into a unique key. That key is recorded in VISID's persistent public index alongside the full declared identity: the title, description, keywords, authorship, usage rights, AI declarations. The intent and significance of the image, permanently anchored to its fingerprint.

Anyone can query that index - by uploading an image or submitting a URL of one encountered anywhere online - and VISID finds its match. If that image was ever stamped, the record surfaces. Who made it. When. What it was for. What rights apply. Even if every piece of embedded metadata was consumed and stripped by every platform it passed through, the fingerprint was never in the metadata layer. It was derived from the image itself. You cannot strip what was never added.

“You cannot strip what was never added.”

This is rehydration. The record is always recoverable.

But the verify layer surfaces something else too. Because images don't exist in isolation - they get resized, reposted, color-adjusted, cropped, and iterated on. VISID tracks this. If an image has been stamped multiple times across its life by different users, verify surfaces its version history - a timeline of how the image evolved, each version anchored to its original line. If a derivative has been created - a similar image stamped separately, close enough in structure to be recognized as related - it enters the family as a connected record. Not the same image, but a descendant from the same visual origin.

At scale, this becomes something genuinely new: a window into how imagery moves and mutates through the public sphere. Which images proliferate. Which get copied. Which originals spawned entire derivative families without ever receiving credit. The verify page is not just a lookup tool - it is the beginning of a provenance layer for visual culture that has never existed before.

Where This Leads

But this is the beginning.

As images accumulate in the VISID index, something else begins to emerge - a living record of visual culture that doesn't forget. Who made what. When. What it meant. Which creators shaped the aesthetic languages that AI is now extrapolating from, often without credit, often without trace.

We are at an inflection point. AI is producing imagery at a scale no individual creator can compete with. In that world, the scarce thing isn't imagery. There will always be more imagery. The scarce thing is authorship, provenance, and meaning. The ability to say - this idea originated here, with this person, at this moment in time.

“The scarce thing isn't imagery. There will always be more imagery. The scarce thing is authorship.”

Without infrastructure to preserve that, we lose it. Not gradually. Quickly.

Every stamp is a declaration. Every declaration is a record that doesn't forget. What that record becomes over time is a different conversation. But it starts here.

If This Resonates

VISID is early. I have limited resources and for the moment I'm intentionally keeping it that way. I'm looking to work closely with a small number of companies where imagery is genuinely seen as an asset, not a commodity. Where the brand is in the image, not just around it.

If that's you - if you're producing high-equity imagery and the idea of giving it a persistent identity resonates - I'd like to talk.

VISID gives images a declared identity before they leave your hands - authorship, intent, context, and rights embedded at the moment of creation. Learn more at visid.app.