How do you write a homepage a machine can understand?

If you want AI systems to recommend your business, your homepage needs to state plainly what you do, for whom, in the first paragraph a machine reads. Not a value proposition. Not a promise. A description.

This sounds obvious enough to skip. It is also the single most common failure I see, and I found it on my own site.

What does a machine actually read?

When an AI system encounters your site — whether during training or when browsing to answer a live question — it is trying to do something narrow: work out what kind of business this is, so it can decide whether to mention you when someone asks about that category.

It reads the top of the page first. Title, meta description, headline, opening paragraph. That is where it expects to find a claim it can attach to your name.

A human visitor brings context. They arrived from somewhere, they have a problem in mind, and they will forgive a clever headline because they can infer the rest. A model has no context and no patience. If the top of your page does not contain a statement of what you do, the model does not infer it — it simply comes away with nothing specific enough to repeat.

What should the first sentence do?

Bind a name to a description. The pattern is unglamorous and it works:

[Company] helps [specific audience] [do specific thing].

Three parts, all necessary. The name gives the model an entity. The audience narrows the category. The thing is what you actually do.

Compare:

We turn complexity into clarity.

against

Northwind helps independent pharmacies manage inventory and regulatory reporting.

The first could be any consultancy, agency or software company on earth. The second is categorisable in one pass, and if someone asks an assistant for pharmacy inventory software, the second business is a candidate and the first is not.

You do not have to lead with this. Keep your headline — headlines earn attention and that matters for humans. Just make sure the sentence underneath it is the plain one.

Should you write "we help" or "[Company] helps"?

Third person is usually stronger, for a reason that is easy to miss: "we" is a pronoun with no referent for a machine reading a fragment. "Northwind helps…" binds the description directly to the entity name, which is the thing the model needs to associate with your category.

I learned this from the wrong side. Zapheron's homepage was rewritten to open with "Zapheron helps media and technology organisations…" — textbook. Our own visibility scoring tool still marked it zero.

The tool was wrong. Its check matched four literal phrases — "we are", "we help", "we provide", "our mission" — and third-person phrasing failed it. So the scorer was penalising the better construction. We fixed the check to accept any subject followed by a describing verb, which is what it should have done from the start.

Worth noting because the same bias exists in plenty of guidance: first person is conventional, not optimal. Write whichever reads well for humans, but know that naming yourself costs nothing and helps.

What belongs above the fold?

In rough priority:

1. A title tag that names the business and the category, not a slogan.
2. A meta description of 50–160 characters that states what you do. Longer gets truncated; shorter wastes the slot.
3. One H1, which can be your headline.
4. A plain description sentence immediately beneath it.
5. Enough specificity to be wrong. "Enterprise solutions" is safe and useless. Naming an industry, a customer type or a problem is what makes you findable.

That last one is where most sites lose. Vague copy feels inclusive — it does not exclude any potential customer. But a description that could apply to anyone gives a model no reason to name you for anything.

A worked example

Zapheron's homepage opened with a line I still like:

The work before the AI work.

Underneath it, the paragraph described a philosophy: strategy, structure and visibility, the things that decide whether AI succeeds long before you pick a tool. Good writing for someone who already knows the category. To a machine, no claim at all — it never said what the company does or who it does it for.

The check for a clear business description scored 0 out of 7.

The fix was one clause. The headline stayed. The paragraph now begins:

Zapheron helps media and technology organisations get strategy, structure and AI visibility right — the things that decide whether AI succeeds, long before you pick a tool.

Same voice, same headline, one added statement of fact. That change, plus trimming an over-length meta description, moved the site's visibility score from 83 to 88, and fixing the scorer's own bug took it to 93.

Where to start

Open your homepage and read the first fifty words as if you had never heard of the company. If they do not tell you what the business does and who it serves, that is the change to make this week — before schema markup, before content strategy, before anything else on the list.

Our AI Visibility Scorer runs this check along with twenty others and takes a few seconds. For the wider picture of what drives AI recommendations, start with what generative engine optimisation actually is.