getting found by ai

If you want your business to get found by ChatGPT, Gemini or Google’s AI summaries, the honest answer comes in two parts. The loss of clickthrough traffic is real and measurable: on a Google search with an AI summary, 8 percent of users still click a regular search result, against 15 percent without one (Pew Research Center, 2025). At the same time, getting found inside AI answers is still a tiny channel: in May 2026, ChatGPT accounted for 0.32 percent of all global referral traffic (SE Ranking, 2026). What demonstrably helps is largely classic SEO plus citable, answer-first content, and none of it requires a separate GEO package or an AI findability subscription.

Those two truths get glued together in the market all the time. The proven problem, falling clicks from Google, is used to sell an unproven solution: optimization for an AI channel that delivers less than half a percent of the world’s web traffic. That is the gold rush. In this essay I separate what the numbers carry from what the sales brochure promises, the way I did earlier with the ai failure numbers, and I end with a priority list for this month.

how much website traffic does ai in google really cost you?

A lot, and it can be measured: as soon as Google shows an AI summary above the results, the chance that someone still clicks through to a website roughly halves. Pew Research tracked the search behavior of 900 Americans and saw the clickthrough rate on regular search results drop from 15 to 8 percent whenever an AI summary was present. A source cited inside that summary got a click from 1 percent. And 26 percent of sessions ended right after the summary, against 16 percent without one (Pew Research Center, 2025).

At the big publishers you can see where that leads. Business Insider watched its organic search traffic fall 55 percent between April 2022 and April 2025 and cut a fifth of its jobs in May 2025; at The New York Times, search’s share of total site traffic slid from 44 to 37 percent (AdExchanger, 2025). Study platform Chegg reported 49 percent less traffic from non-subscribers in a single year, though that figure comes from its own filings in a lawsuit against Google and is therefore self-reported (Search Engine Journal, 2026).

How big the hit is exactly depends on who is measuring. Pew arrives at 8 versus 15 percent, Seer Interactive measured the organic clickthrough rate falling from 1.6 to 0.6 percent (AdExchanger, 2025), and analyses by Ahrefs and Amsive put the decline for the top search result at 34.5 and 15.5 percent respectively (LinkedIn, 2025). One hard zero-click number does not exist; the direction is the same in every serious measurement.

And the Netherlands? Independent Dutch figures, from Statistics Netherlands (CBS) or IAB Netherlands for instance, do not exist yet. What circulates comes from agencies: marketing agency Envoker estimated that by mid 2025 AI Overviews already appeared on some 14 to 15 percent of Dutch desktop searches, putting the Netherlands ahead of most of the EU (Envoker, 2025). Take that as direction, not as a hard number.

how often does chatgpt recommend a business like yours?

Rarely. Research by software company SOCi across more than 350,000 locations of 2,751 brands found that a local business was recommended by ChatGPT, Gemini or Perplexity in 1.2 percent of relevant cases on average, against 35.9 percent visibility in Google’s classic Local 3-Pack (SOCi, 2026). Gemini did best at 11 percent, because it leans on Google Maps data. The same research found that business details in ChatGPT and Perplexity were correct in only about 68 percent of cases (Search Engine Land, 2026). This is American vendor research; read it as an indication of the proportions, not as Dutch market data.

The source of the new traffic stays thin as well. According to clickstream data from software company Similarweb, 2.8 percent of ChatGPT answers contained a source citation in August 2025, up from 0.6 percent in January of that year (Similarweb, 2026). After a product change in May 2026, clickable brand links inside the answers, ChatGPT referral traffic jumped 157.7 percent in a single week. That sounds spectacular; it grew from nearly zero. Globally, ChatGPT sat at 0.32 percent of all referral traffic in May 2026, and within that small AI segment it is losing ground to Gemini and other competitors on top of it (SE Ranking, 2026). One bright spot from the same vendor data: visitors who do arrive via ChatGPT convert at 7.1 percent, nearly on par with paid search traffic (Similarweb, 2026).

The sober conclusion: AI referral traffic is growing fast and stays marginal for now. Track it, and do not hang a budget or a revenue target on it yet.

what demonstrably works for getting found by ai?

Four things have evidence or solid logic behind them: answer-first writing, numbers and sources in your content, structured data (schema.org), and a consistent entity with matching business details and up-to-date reviews.

The only peer-reviewed research on GEO comes from Princeton, Georgia Tech and partners, presented at KDD 2024. Across 10,000 test queries, adding sources, quotations and statistics raised visibility in AI answers by 30 to 40 percent on the researchers’ own metric, with gains of up to 37 percent in a validation on Perplexity; keyword stuffing did next to nothing (Aggarwal et al., 2024). The caveat belongs with it: this measures visibility in a test setup. It is no proof that a real business gains 30 percent in revenue.

For local businesses there is a second signal. The American Whitespark report on local ranking factors introduced a separate category for AI visibility in 2026, and within it the content on your own website carries the most weight (24 percent), against 12 percent for the Google Business Profile; in the classic local pack that ratio runs the other way (Whitespark via Localo, 2026). Loosely translated: for AI answers your website matters again, even if your work is mostly local.

The through line is always the same. AI systems cite pages that answer a question directly, verifiably and with sources, from a business whose name, address and offering are the same everywhere. That is largely the SEO and content work that has been sensible for years. I apply it here myself: this essay opens with the full answer, every number has a source, and at the bottom sits an FAQ with matching schema. That took an afternoon of writing and zero subscriptions.

what you do not need to buy

Top of the list: llms.txt, the text file that is supposed to show AI crawlers the way. The honest story about it is short. It costs nothing, it does no harm, and there is no evidence it does anything. Google’s John Mueller publicly compared llms.txt to the keywords meta tag from the nineties and noted that no AI service has said it uses the file; Google does not use it either (Search Engine Journal, 2025). Anyone offering you an llms.txt implementation as a paid service is selling you a file nobody reads.

The same goes for AI visibility dashboards and ‘AI rankings’ as a standalone subscription. Knowing what ChatGPT says about your business is sensible, and for that you open ChatGPT and ask it, free of charge, every quarter. Meanwhile the GEO industry has landed commercially in the Netherlands too: in May 2026 a Dutch service for AI findability launched with quick scans and fix packages, announced via a submitted press release (Emerce, 2026). That says nothing about that one provider; it says a great deal about the sales wave that is coming.

Concretely, this is what you do not need to buy this year: a paid llms.txt implementation, a monthly AI visibility subscription, a separate GEO audit next to your regular SEO work, or tooling that tracks your position in ChatGPT daily for a channel that delivers less than half a percent of your traffic. Buying tooling first and building the foundation afterwards is also the exact pattern that strands most AI projects between demo and daily practice; I wrote about that in the production gap.

the priority list: what you do this month

Everything below you can do yourself, or with the web developer you already have.

1. Measure your baseline in Google Search Console. Look for pages where impressions hold steady while clicks fall: that is where an AI summary is already giving the answer away.
2. Rewrite your five most important pages answer-first. The first paragraph answers the visitor’s question in full; explanation and nuance follow after.
3. Add numbers, sources or quotations to those pages. This is the tactic with the strongest research behind it (Aggarwal et al., 2024).
4. Get your structured data in order: Organization or LocalBusiness for your company, Article for articles, FAQPage wherever you answer questions.
5. Make your entity consistent: the same name, address and phone number on every listing, an up-to-date Google Business Profile, and respond to reviews.
6. Do a free baseline check: ask ChatGPT, Gemini and Perplexity what they know about your business and which sources they name. Write down the outcome and repeat it every quarter.
7. Reduce your dependence on Google: an email list, returning customers, partnerships. Traffic you own does not fall with the next CTR chart.

And what you skip: adding llms.txt is fine (five minutes of work), just expect nothing from it. No paid monitoring and no agency until steps 1 through 5 are demonstrably done.

My position is this: the loss of Google clicks is the real problem and deserves your attention now; getting found by AI assistants is hygiene you fold into content work that needs doing anyway. Anyone promising you a GEO breakthrough today is selling a solution to the wrong problem.