The old model was simple enough to draw on a napkin: search, scan the results, open a few sites, call one. It is no longer accurate, and marketing plans still built on it fail in ways that are invisible in reporting.
One: Google AI Overviews
A generated answer above the results, citing a small set of sources. Rewards pages that state the answer plainly and companies that search engines can already identify with confidence. Companies disappear here by burying the substance.
Two: Google AI Mode
A conversational surface where the follow-up questions — what does it cost, how long does an engagement take, do we qualify for this service — happen inside Google rather than on a company's website. Being in the retrieved source pool is the entire game.
Three: assistants
ChatGPT, Gemini and Perplexity, used as a first stop. Asked to recommend a provider, they typically name two or three companies or point at a directory. Entity clarity and third-party citation presence appear to govern who is named; output varies by phrasing and session, which is why sampling rather than a single check is the only sound way to observe it.
Four: Maps and the local pack
Still the surface with the clearest commercial intent, and still governed by proximity, category accuracy, profile completeness and reviews. It is also the data that grounds most place-specific prompts on the other surfaces.
Five: classic organic results
Not gone. It remains the pool nearly everything above retrieves from, and for many high-value service queries it is still where the click happens. A company that abandons it in favor of chasing AI visibility has misread the mechanism.
What this means for planning
The five surfaces share most of their inputs and differ in emphasis. That is good news for a company with a limited budget: the work compounds. What has to change is measurement. A program reported only on rankings and sessions is describing one surface and calling it the market.