How a small company gets on the first page of Google

8 min read · Updated 21 August 2026

Start with what Google actually promises.

Nothing. That is not a rhetorical opening — it is in the documentation. Google Search Essentials sets out the technical requirements, the spam policies and the best practices, and then says this:

So any supplier offering guaranteed first-page placement is contradicting the only party in a position to know. That single sentence is the most useful thing an owner can carry into a conversation with an agency, and it costs nothing to check.

What follows is separated deliberately into three categories, because they carry very different weight: what Google documents, what the industry infers, and what is being sold and does nothing.

What Google documents.

This part is narrower than most people expect. Google's own framing is that the technical bar is low — in its words, there are “actually very few technical things you need to do to a web page; most sites pass the technical requirements without even realising it.”

That is close to the whole of what is actually documented. Everything past this point is inference.

What the industry infers.

Google does not publish its ranking factors, so the widely repeated lists are assembled from observation, correlation studies, statements by Google staff, and one substantial documented leak. They are not worthless and they are not documentation, and the difference matters when someone quotes them at you as fact.

The recurring themes are consistent enough to be worth acting on: content that demonstrably answers the query, links earned from relevant sites, a site that covers its subject in depth rather than in one page, and genuine updates rather than cosmetic ones.

What is being sold that does nothing.

The clearest current example is a file called llms.txt, widely recommended as the way to be visible to AI search. Google addresses it directly in its own guidance on generative AI features:

The same guidance is blunt about the broader category. Google states that its generative AI features are rooted in the core Search ranking and quality systems, that no special schema.org markup is required for them, and dismisses tactics like content chunking and keyword-focused rewrites. In other words, on Google, optimising for AI answers is optimising for search. The new vocabulary — generative engine optimisation, answer engine optimisation, and several other labels with no agreed definition between them — describes a real shift in how results are presented, but not a separate technical discipline you can be sold.

A word about leaning on Google’s own documentation, since Google owns the algorithm and is hardly disinterested. Trust it on mechanism: whether Search reads a particular file is something only Google can answer, and it gains nothing by inventing extra work for itself. Treat its strategic advice more carefully — “create helpful, people-first content” is unfalsifiable, and unpredictable organic results do make paid results more attractive.

That caution applies to us as well. We build websites for a living, and it would be commercially convenient to tell you there is a new specialism only we understand. There is not, at least not yet. Almost everything in this article can be checked in a browser in ten minutes, and you should.

The route that actually works for a small company.

Here is the part most quotes leave out. A small firm competing for the broad category term is competing against organisations with twenty years of accumulated links and a marketing department. That contest is usually not winnable and rarely worth entering.

The specific question is a different market. Someone typing a broad industry term is browsing. Someone typing a precise question with a number or a condition in it is much closer to buying, the competition for it is thinner, and a small company can genuinely win it.

We can show the difference on our own site, which is the only one we are entitled to publish positions for. Both figures are from Google Search Console over the same twenty days:

Same site, same authority, same twenty days. The page answering a precise question sits on the first page; the page competing for the category term sits on the third. That gap is not a quirk of our site — it is the shape of the opportunity for any small company, and it points at what to write.

The one split that tells you whether any of it is working.

If someone reports on your search performance and gives you a single number for traffic or impressions, ask them to split it in two. It is the fastest way to tell a real result from a flattering one.

Both appear in the same total, and only one of them is discovery. A company whose name is becoming better known through ordinary trading will see its search numbers rise whether or not anybody has touched the website, and that rise can be presented as a result. Splitting the two makes it obvious.

There are two ways this split can look wrong, and they need different fixes. Mostly branded means the website is serving people who already found you some other way — it is working as a business card, not as a source of enquiries. Plenty of non-branded impressions but almost no clicks usually means position: appearing on the second or third page counts as an impression, and almost nobody clicks past the first. Both are useful to know and neither is visible in a combined figure.

Search Console shows the split for free on your own site, and you do not need an agency to read it.

What to do on Monday.

Write down the questions a buyer asks you in the week before they decide. Not the ones you wish they asked — the real ones, including the awkward ones about price, scope and what is not included. Then check how many of them your website answers on the page, rather than deferring to a phone call.

In our experience that number is lower than the owner expects, and closing the gap is both the cheapest search work available and the most likely to produce an enquiry, because the person arriving has already had their question answered and is arriving to act on it.

It is also the version of this work that survives every algorithm change, because it is not a tactic. It is just telling people what they wanted to know before they had to ask.