Getting found by AI is now somebody's job because an assistant stands between the buyer and the business. The assistant reads the site and decides whether to name the business at all. It reads a layer of the site that no team was hired to look after. In most companies the work falls between marketing, IT and whoever writes the product copy.
For twenty years, being found meant getting a person to click a link. Now the buyer asks a question and the assistant writes one answer. The business is named in that answer or it is left out, and the buyer rarely asks again.
What changed when assistants started answering?
Search engines gave the buyer a list and let the buyer choose. An assistant does the choosing first. It reads what it can parse about each business, compares what it found and names a few. A business that the assistant could not read never reaches the comparison.
That shift moves the decision earlier, onto the site itself. Design still matters once a person arrives, but the layer underneath decides whether the assistant sends them.
Why do machines read a site differently from people?
A person reads the design. A model reads the entities, which are the structured facts underneath the page. They say who the business is, what each product is, what it costs and whether it is in stock. Most teams keep improving the design and never check the facts underneath, since nothing on screen shows them.
In our June 2026 study of 216 Shopify storefronts, the median store scored 46 out of 100. Only 27 of the 216 passed all five layers of our score at half marks or better. The weakest layer was Entity Integrity, the brand and product data machines read, and 72% of stores failed it.
Can AI legibility be measured?
We measure AI legibility as a number. Our AI Visibility Score combines five measures into one score from 0 to 100. They are the Front Door, Entity Integrity, AI Readability, Mesh Integrity and Authority Signal. The AI Legibility Framework v1.0 defines each one.
Every figure Visibility Mesh publishes traces to a live scan, with its population and date. A claim that cannot be traced to data does not get published. That rule costs us some good headlines, and we have made our peace with it.
Why is this bigger than ecommerce?
The same questions apply to a law firm, a clinic, a university or a manufacturer. Anywhere an assistant decides whether an organization is clear enough to mention, it reads the same kind of layer. We started with online stores because they are easy to measure at scale. Our July 2026 study of 303 non ecommerce websites found the same weak spot in structured data.
Why does acting early cost less?
In our benchmark of 245 Shopify storefronts on August 22, 2026, most stores sat in the middle bands and no store reached the Leading band. Since nobody holds the top yet, a business that fixes its readable layer this year has very few stores ahead of it.
Whose job is it, and how does a business start?
The job has parts that belong to different people. Marketing owns what the site claims, and IT owns whether crawlers can get in. Content owns whether answers are written plainly. The founder decides which facts the business stands behind. Who owns AI visibility splits the work role by role.
A business can start this week with five steps.
- Name one person who owns AI visibility, even if the fixes are done by others.
- Run the free scan and save the scorecard as the baseline.
- Ask ChatGPT and Perplexity "What is [your business], and what does it do?" and write down what they say.
- Give each finding on the scorecard to the role that owns it, with a date.
- Scan again after the fixes and compare the two scorecards layer by layer.
What does this look like in a real kind of business?
A furniture maker on Shopify has a marketing manager, an outside developer and a founder who writes the product descriptions. The scan shows a duplicate Organization block, product dimensions written only inside images, and a robots.txt rule an old app added. The marketing manager thinks the developer handles markup, and the developer thinks marketing does.
Once one person owns AI visibility, the three findings get three names and three dates. The developer removes the robots rule, the marketing manager removes the duplicate block, and the founder types the dimensions into each description. None of the three fixes was hard, and none had been done because nobody owned them.
Questions leaders ask
Why is AI visibility a separate job from SEO?
SEO grew up around ranked lists of links. AI visibility is about whether an assistant can read and trust the structured facts on a site. That work touches marketing, IT and content at once, so it needs one owner who looks across all three.
Does this apply to businesses that do not sell online?
It does. Our July 2026 study of 303 non ecommerce websites found the same weak spot in structured data. Assistants answer questions about law firms, clinics and manufacturers too.
Who should own AI visibility in a small business?
In a small business the founder or the person who runs the website usually owns it. The owner does not need to make every fix, but the owner keeps the scorecard and gives each finding to the right person.
Can anyone guarantee that AI will recommend my business?
Nobody can guarantee that, and we do not. A business can measure whether assistants can read its site. It can then fix what they cannot read, and that part is in its control.
Sources
- The State of AI Visibility 2026, Visibility Mesh, 216 Shopify storefronts scored in June 2026.
- The State of Shopify AI Legibility, Visibility Mesh, 245 storefronts read on August 22, 2026.