Do-it-yourself seriesOnline visibility basics
Visibility basics are the minimum you need in place before people can find your business, and before the AI systems that increasingly answer on their behalf can name you. This isn't about ranking first nationwide. It's about being found by the people already looking for something like you.
I first wrote this series in Dutch, back in 2017. It has now been brought up to date: the same foundation, plus the part that got added since, which is getting found in AI answers.
What are visibility basics?
Having a website isn't visibility. Visibility is somebody with a question ending up at your door: through Google, through Maps, through an AI answer, or because somebody else mentioned you. The foundation is five things.
- Your business details are correct and findable. That starts with your Google Business Profile.
- Your website can be understood by a visitor and by a machine reading it.
- Your text answers real questions with things only you can say.
- Other people mention you. That counts for far more than it used to.
- You measure what can be measured and stay honest about the rest.
That's all. No tricks. When it goes wrong, it's almost never because somebody couldn't do one of these. It's because nobody ever wrote them down together.
Where do you start?
With whichever part is weakest for you, not whichever part sounds most interesting. See which of these you recognize.
| What you see | What it usually comes down to | Read on |
|---|---|---|
| People nearby cannot find you | Your business details aren't the same everywhere | Your Google Business Profile |
| Visitors land on the wrong page | Your site doesn't have a clear structure | Structuring your website |
| You're in Google, but AI never names you | Your text is hard to quote, or a crawler isn't allowed in | Text that AI can cite and AI crawlers |
| You publish a lot and little comes of it | You're writing things ten other people also write | Brand name and mentions |
| You don't know whether any of it works | You're measuring things that can no longer be measured | Measuring honestly |
And before you read on, here are four checks you can run right now, about five minutes for all of them together.
- Search for your own company name in Google. Is everything in the panel on the right correct, opening hours included?
- Open
yourdomain.com/robots.txt. Is there anything in there you don't understand, or nothing at all? - Turn JavaScript off in your browser and reload your most important page. Whatever you can't see then, plenty of crawlers can't see either.
- Ask ChatGPT or Gemini a question you ought to be the answer to. Who gets named?
Whatever comes out of that is your starting point. The rest of this page explains what to do about each part.
What this isn't
This belongs at the top rather than the bottom, because it might save you the rest of the page. These are the foundations for getting found, and getting found is one part of a marketing strategy. This series isn't about advertising, positioning, pricing or your offer itself.
That distinction matters, because it's the most common misdiagnosis I run into. If nobody wants what you're selling, being easier to find won't help you. It only means more people get to say no. Getting found widens the reach of what you already have. It doesn't repair it.
Not sure which of the two your problem is? That's exactly the kind of question I like to look at before anybody spends money. Otherwise start with what I do, or read how I think about SEO and AI visibility.
What changed since 2017?
Enough to rewrite the series. The four that matter most:
- Google My Business is now the Google Business Profile. The separate dashboard was switched off in 2022; you manage the profile from Google Search and Maps instead.
- Universal Analytics stopped in July 2023. GA4 measures differently, and comparing it against your old figures doesn't hold up.
- Measurement became partly estimation. For visitors in the EEA you need consent, and since March 2024 Consent Mode v2 is required if you use Google's advertising services. Part of what you see in your reporting is modeled rather than observed.
- A layer got added. AI Overviews in Google, plus ChatGPT, Perplexity, Copilot and Gemini as starting points. More and more often somebody gets their answer without clicking through.
What hasn't changed: you're better off with a solid foundation than with a trick. That was true in 2017 and it's more true now, because the tricks get noticed faster.
How does searching through AI actually work?
Broadly: the system searches first and writes afterward. It searches an index, a handful of sources come back, and the model turns those into an answer with references. That's called grounding.
The most important conclusion in this whole series follows from that: if you're not in the underlying search index, you can't be in the answer either. Being findable in ordinary search engines still returns by far the most. "AI SEO" is mostly just SEO, plus making sure your text is easy to quote.
You don't have to take my word for it. Google writes it themselves in their guide to generative AI features in Search: "The best practices for SEO continue to be relevant." The reason they give is that the generative search features lean on the existing ranking and quality systems. In other words: it's the same engine.
So be suspicious of anybody telling you otherwise. A lot is being sold right now on claims nobody can back up.
Read how AI search engines find, read and cite your website, and which crawlers to allow, or see what I do under SEO and AI visibility.
Which AI crawlers do you allow?
This is a real decision, and it usually gets made on bad information, because the different robots get treated as one thing. They don't do the same job: some fill a search index, some fetch a page because a user asked for it at that moment, and some collect material to train models on.
| Robot | What for | What blocking it costs you |
|---|---|---|
Googlebot | Google's search index; also feeds AI Overviews | You disappear from Google entirely |
Google-Extended | Training and grounding for Gemini | Nothing in AI Overviews, only in Gemini |
OAI-SearchBot | ChatGPT's search index | No more mentions in ChatGPT search results |
ChatGPT-User | Fetches a page when a user asks for it | Can't be reliably blocked through robots.txt |
GPTBot | Training OpenAI models | Not used for training |
ClaudeBot | Training Anthropic models | Not used for training |
Claude-User | Fetches at a user's request | Your page can't be opened on request |
Claude-SearchBot | Claude's search index | Less visibility in Claude's search results |
PerplexityBot | Perplexity's index | No more mentions in Perplexity |
CCBot | Common Crawl; feeds many datasets | Out of a widely reused corpus |
Applebot-Extended | Training by Apple | Not used for Apple training |
What these robots do is documented by OpenAI, Anthropic, Perplexity and Apple.
One misunderstanding is stubborn enough to call out separately: you can't opt out of AI Overviews through robots.txt. Those are built on Google's ordinary search index. Blocking Google-Extended is about Gemini, not about whether you appear in AI Overviews. Google says so in as many words in their list of crawlers: "Google-Extended does not impact a site's inclusion in Google Search nor is it used as a ranking signal in Google Search." If you genuinely want out, the only levers are nosnippet and max-snippet, and those cut your ordinary search snippet too.
My own choice on this website: everything allowed. There's nothing here that needs protecting, and I'd have a hard time writing about AI visibility while holding the door shut.
Don't know what's in your robots.txt right now, or what it's shutting out? Send me your website and I'll tell you which robots are coming in and what that costs you.
How do you write text an AI can cite?
By putting the answer first. These systems retrieve passages, not whole pages. A passage that makes sense on its own gets quoted; a passage where the answer shows up in the fourth paragraph doesn't.
- Answer in the first sentences under every heading. Explanation, nuance and examples come after.
- Make your headings the questions somebody would actually ask.
- One claim per paragraph, with the evidence in that same paragraph.
- Be specific. A year, an amount, a name. "Many companies find that" is not information.
Being unique is no longer a goal in itself. That was the advice in 2017 and it's outdated: unique text is the minimum, not an advantage. What sets you apart is what you've been through yourself and can back up.
What does structured data get you, and what doesn't it?
Structured data, usually schema.org in JSON-LD, tells a machine literally what's on the page: this is a company, this is a person, this is an article with this date and this author.
What it demonstrably does: it lets Google show certain presentations, such as breadcrumbs, and it helps distinguish you as an entity from somebody with the same name. That second part is handled well with sameAs references to your other profiles.
What it doesn't do: no AI search provider has documented that JSON-LD counts toward which sources get chosen. Anybody telling you schema.org gets you into ChatGPT is selling you something.
Watch out for two things that no longer hold, too. Google announced in August 2023 that FAQ results would only be shown for authoritative government and health sites and that HowTo results would go; the latter aren't shown at all any more. So that markup on your site no longer earns a rich result. You don't have to remove it, but don't expect anything from it.
Source: Google Search Central on the changes to HowTo and FAQ results.
What do you do with your Google Business Profile?
This was part one in 2017 and it's still where you begin, certainly if your customers come from nearby. Your profile decides whether you show up in Maps and in the local block, and it's one of the sources AI answers pull business details from.
Make sure the name, address, phone number and opening hours are exactly the same everywhere: on your profile, on your website and anywhere else you're listed. Differences between them are the most common reason a system can't be sure it's dealing with the same business.
Ask for reviews, and respond to them. Not because some number crosses a magic threshold, but because a recent review with a specific story says something you can't credibly write about yourself.
How much a review counts varies by market. In Norway I see them carry more influence than in the Netherlands, and that isn't the only difference: read why it's easier to get found in Norway than in the Netherlands.
How do you structure your website?
This was part two. The core hasn't changed: every page has one clear subject, and from your homepage you can reach anything in a few steps.
- One subject per page. Two services on one page means you get found well for neither.
- Headings in order. One h1, h2s below it, h3s only below those. Don't skip levels.
- Link with descriptive text. "Read more" tells nobody anything. "How to set up your Google Business Profile" does.
- Keep your URLs readable and leave them alone. Every change costs you the value of the links pointing at them.
That last one is a mistake I made myself. This series has had three different URL structures over the years. Every move cost part of the value it had built up.
If you do have to move, make sure every old URL redirects to the page that genuinely replaces it. That work falls under SEO and AI visibility.
Why does your brand name matter more than your keywords?
Because these systems combine sources rather than count links. If your company name turns up in several independent places in the same context (a trade publication, a customer story, an industry association, a podcast), it becomes more likely you get named when somebody asks about that subject.
That holds for mentions without a link, too. That's a real difference from ten years ago, when a mention without a link got you almost nothing.
In practice that means: be genuinely good at something, and make sure it shows outside your own website. That's slower than putting keywords on a page, and it's a great deal harder to copy.
What can you still measure honestly?
Less than your reporting suggests, and that's no reason to stop measuring. It's a reason to read the numbers differently.
Visitors who don't give consent aren't measured; part of that gets estimated and filled in. And the path from first contact to signed work rarely runs through a single channel.
Traffic from AI answers is hard to recognize on top of that, because plenty of AI clients don't pass on where somebody came from. How large that share is, I don't know, and I haven't seen a measurement of it I'd trust. So assume you're underestimating it rather than overestimating it, but don't put a percentage on it.
What does work: look at the things you're certain about. How many inquiries did you get, from what kind of companies, and what did they come to? Just ask new clients how they found you. That answer is imprecise, but it's honestly imprecise, and that's worth more than a dashboard suggesting a certainty that isn't there.
Measurement is also exactly where advertising goes wrong. I write more about that under performance marketing.
What's proven and what isn't?
Less than is claimed. Below is what you can reasonably assume about each statement, so you can judge for yourself what somebody is selling you.
| Claim | Status | Why |
|---|---|---|
| If a crawler can't fetch your page, you can't be cited | Demonstrably true | You handle it in robots.txt and you can check it per robot |
| AI answers lean on the ordinary search index | Well supported | Searching comes before writing; published practical tests point the same way |
| Putting the answer first helps | Well supported | Retrieval happens at passage level, not per whole page |
| Schema.org gets you into AI answers | Overstated | No provider documents JSON-LD as a signal in choosing sources |
| FAQ and HowTo markup earns a rich result | No longer true | Google limited FAQ to government and health sites; HowTo results aren't shown any more |
llms.txt improves your visibility | Refuted | Google says explicitly that it ignores such files and that they neither help nor harm your visibility |
| AI visibility scores measure your position | Overstated | A sample from a system that doesn't give the same answer every time |
The bottom four rows currently cost the most money for work that returns nothing. That's no reason to leave schema.org out, because it genuinely does something for how Google understands you. It is a reason not to buy it as an AI solution.
On llms.txt Google is by now unambiguous: "You don't need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search." Which is why there's no llms.txt on this website. See Google's guide to generative AI features.
Want to go through this together?
Tell me briefly what your business does and where you're stuck. I'll tell you honestly whether getting found is your problem or whether something else is going on.
Dominique@framvekst.noYou'll hear back within three business days.
See my contact details