Salesforce asked 3,450 commerce professionals a question this spring that sounds like plumbing rather than strategy: is your customer data fully unified across sales, service, marketing, and commerce? Twenty-seven percent said yes.
Hold onto that number, because in the same body of research, agentic search grew 200% year over year. Agentic search means a shopping journey that starts inside an AI chat instead of a search bar. The entry point to the funnel moved, and most of the companies that spent a decade optimizing the old entry point are not ready for the new one.
The short version: AI search for small business is not a channel you buy your way into. AI assistants recommend businesses whose facts are consistent and retrievable, and the largest companies surveyed are the ones admitting they cannot keep their own facts straight. A single-location business with one phone number, one price list, and one set of hours has a structural advantage here that it never had in paid search. The work is unglamorous, and most of it is free.
What actually changed in how customers find you?
Two independent datasets point the same direction. Salesforce’s commerce research, published in late July and covering data through May 2026, combines traffic analysis across more than 1.5 billion shoppers in 37 countries with surveys of 3,450 commerce professionals and 4,689 consumers. Referrals from AI chats grew between 150% and 428% year over year in every quarter measured, while overall traffic grew in the single to low double digits. Agentic search itself rose 200%.
Adobe Analytics, working from a different vantage point, tracked more than a trillion visits to United States retail sites. In May 2026, traffic arriving from AI sources was up 138% year over year, and up 1,324% since Adobe began tracking it in October 2024. The more interesting figure is what those visitors did after arriving: they converted at a rate 54% better than non-AI traffic, spent 53% more time on site, and viewed 23% more pages.
A year earlier, AI-referred visitors converted at roughly half the rate of everyone else. That reversal is the actual news. People arriving from an AI conversation used to be browsing. Now they are arriving pre-qualified, because the assistant did the comparison work before the click.
Why does 27% matter more than 200%?
Growth numbers get the headline. The readiness numbers are where the opportunity sits.
Alongside that 27% on unified data, the Salesforce survey found 32% of organizations have defined success metrics for AI, 41% of multichannel organizations report inconsistent pricing and promotions across their own channels, and 40% report problems keeping inventory synchronized in real time. Twenty-eight percent are currently using agentic AI, with another 44% planning adoption within six months, and 86% expecting large language models to be essential to product discovery within a year.
Read those together and a specific picture emerges. Large organizations understand exactly what is happening and largely cannot execute on it, because the blocker is not insight or budget. It is that their prices, hours, inventory, and product facts live in six systems that disagree with each other.
Caila Schwartz, who leads shopper insights for agentic commerce at Salesforce, put it plainly: “A customer can adopt AI overnight. A business has more to get right: trusted data, connected systems, teams ready to use them.”
Does AI search favor big brands or small ones?
Here is the part worth sitting with, because it runs against a decade of instinct.
In paid search, the binding constraint was money. If a national competitor decided to own your keyword, they owned it, and no amount of cleverness closed a budget gap of that size. Small businesses lost that fight on arithmetic.
In AI-mediated recommendation, the binding constraint is coherence. An assistant answering “who does emergency plumbing near me on a Sunday” is assembling a recommendation from whatever it can find and reconcile about you: your site, your Google Business Profile, directory listings, review platforms. When those sources contradict each other on hours, service area, or pricing, the assistant has a reliability problem, and it tends to route around reliability problems.
Reconciling four sources is an afternoon of work. Reconciling six enterprise systems across 40 markets is a multi-year program with a steering committee. The 27% figure is enterprises reporting, in their own words, that they have not finished it.
This is not a claim that small businesses will out-rank national chains in AI answers. Brand recognition, review volume, and genuine authority still carry enormous weight, and they always will. The narrower and more useful claim is that this particular new requirement, consistent and machine-readable facts, is one of the few competitive requirements in modern marketing where being small is an advantage rather than a handicap. Take the wins that are actually available.
What should a small business do about it this month?
Nothing here requires a new hire, a new platform, or a consultant. It requires an afternoon and some stubbornness.
Start by asking an AI assistant about your own business the way a customer would. Not “tell me about [your business name],” which flatters you, but the question a stranger asks: the best option in your city for the thing you sell. Note what it says about you, what it gets wrong, and who it names instead.
Then open every place your business facts live, side by side. Your website, your Google Business Profile, Yelp, industry directories, your social profiles. Check hours, service area, phone number, business name formatting, and what you actually sell. Contradictions are common and boring, and each one is a small reason for an assistant to prefer someone else.
Write the answers down where a machine can read them. The questions customers actually ask, answered in plain sentences on your own pages, do more work than a paragraph of marketing prose. If you charge extra for emergency calls after 6pm, say the number. If you do not serve a neighboring county, say so. Specificity is what gets retrieved.
Then check whether you can respond when the qualified visitor arrives. Adobe’s data says these people convert better, which only matters if someone answers. Our guide to AI customer service for small business covers the leaks that swallow those leads, and our practical playbook for AI in small business marketing covers the tooling around it.
One caution worth carrying. As AI agents increasingly act on customers’ behalf, the question of who answers for what an agent does on your site is being settled right now, and a recent federal appeals ruling placed more of that responsibility on site owners than most expected. Being findable by agents and being ready for them are two different projects.
Frequently Asked Questions
Is AI search actually replacing Google for small business customers?
Not replacing it, but supplementing it faster than most owners realize. Salesforce found agentic search up 200% year over year, with AI chat referrals growing between 150% and 428% in every quarter measured, while total traffic grew in the single to low double digits. Traditional search remains far larger in absolute volume. The shift that matters is where the journey begins, because the assistant increasingly decides which handful of businesses the customer ever sees.
Do I need to pay for a tool to show up in AI answers?
No, and be skeptical of anyone selling guaranteed placement in AI results, since no such guarantee exists to sell. The inputs assistants draw on are things you already control and mostly already have: your website content, your Google Business Profile, your directory listings, and your reviews. The work is making them accurate, specific, and consistent with each other, which costs time rather than money.
Why would AI traffic convert better than search traffic?
Because the filtering happened before the click. A customer who asked an assistant to compare options and then visits you has already been through a comparison step that used to happen across several browser tabs, or not at all. Adobe measured that traffic converting 54% better than non-AI traffic in May 2026, spending 53% more time on site and viewing 23% more pages, a complete reversal from a year earlier when the same traffic converted at about half the rate.
How do I check what AI assistants currently say about my business?
Ask them directly, in a fresh conversation, using the phrasing a customer would use rather than your business name. Try two or three assistants, since they draw on different sources and will disagree with each other. Write down the factual errors and which competitors get named, then trace each error back to whichever listing or page is feeding it, because that traceable source is the thing you can actually go and fix.
We are curious about one thing in particular: when you asked an assistant about your own line of work, did it name you, and if not, what did it get wrong? Tell us in the comments.
