Your Attribution Model Cannot See Where the Decision Gets Made

The gap is not new, it is newly large

Marketers have always known that dark social, word of mouth and offline conversation influence purchase. The difference now is scale and permanence.

A Reddit thread from 2023 is still being read, still ranking, and still feeding AI models that summarise your category to prospects. A YouTube comparison video keeps working for years. When someone asks an AI assistant whether your platform is trustworthy, the answer is assembled from sources you may never have audited, and the prospect receives it as a neutral verdict rather than as marketing.

That last point is what makes this different from every previous channel shift. Traditional channels are recognised as advertising, and buyers discount them accordingly. A recommendation surfaced inside an AI answer or upvoted in a community carries the weight of independent verification. It is more persuasive precisely because it does not look like you paid for it.

Why this hits regulated sectors hardest

In iGaming, finance, legal and healthcare, the purchase decision carries risk. Nobody deposits funds, switches broker or instructs a solicitor on the strength of a paid search ad alone. They go looking for evidence that you are safe.

That verification behaviour concentrates in exactly the places attribution cannot see. Casino affiliate forums. Trustpilot threads. Subreddits where regulated-industry consumers compare licences, withdrawal times and complaint handling. Community answers on Reddit that get pulled wholesale into AI Overviews.

There is a second problem specific to regulated brands. Compliance restricts what you can say about yourself, but it does not restrict what other people say about you. Communities and AI assistants will fill that vacuum whether or not you participate. If your brand is absent from those conversations, the absence itself becomes a signal, and competitors who have invested there own the answer to “is this operator legitimate”.

We ran Brand Radar analysis across a set of client and prospect brands earlier this year. The pattern repeated: strong Google rankings, meaningful organic traffic, and close to zero share of voice inside ChatGPT and Google AI Overviews for the category questions their buyers were actually asking. Good SEO performance masking a discovery problem.

What to do about it

Audit what the machines say about you

Ask ChatGPT, Gemini and Perplexity the questions your buyers ask. Not “tell me about [brand]”, but the real ones: best licensed sportsbook for UK players, safest crypto exchange for beginners, which firms handle industrial disease claims. Record which brands appear, which sources get cited, and where you sit. This takes an afternoon, and it is the single most useful hour of work available to most marketing teams right now.

Find the citation sources, then earn placement in them

AI answers are assembled from a surprisingly narrow set of sources per query. Once you know which domains and threads feed the answers in your category, the target list stops being theoretical. Reddit and trade press consistently punch above their weight here, which is why community strategy has moved from a nice-to-have into core visibility work.

Stop treating last click as truth.

You will not get clean attribution for community and AI-assisted discovery, and waiting for a tool to solve it is a way of doing nothing for two years. Use self-reported attribution at the point of conversion, a single “how did you first hear about us” field, and combine it with branded search volume as a proxy for upstream demand creation. Imperfect signal beats confident nonsense.

Watch branded search as an outcome metric, not a starting point

If community and AI visibility work, branded search volume rises before revenue does. That lag is your evidence base. It is also the argument that keeps the budget alive when someone asks why you are spending money on channels with no direct conversion path.

The uncomfortable part

Most of this work is unattributable by design. You are investing in surfaces that will not report back to you, on the basis that buyers are being persuaded there. That is a harder internal sell than a paid search line item with a CPA next to it.

But the alternative is optimising towards the point in the journey where the decision has already been made, and calling that growth. The brand that owns the answer when a prospect asks an AI assistant whether you are worth trusting has already won, months before your attribution model notices anything happened at all.