AI recommendations are reshaping small business traffic in two ways at once. Google answers more questions on the results page, so fewer searches send anyone to a website: SparkToro measured 68.01% of US Google searches ending without a click in early 2026. And AI assistants have become a place where people choose local businesses: in BrightLocal's 2026 survey, 45% of US consumers had used AI tools for local business recommendations, up from 6% in 2025. The visits that do arrive from assistants are fewer, but the people behind them are further along, and many of them come to your site to check a recommendation rather than to discover you.
What actually changed
Picture a plumber in a mid-sized town. On a results page, a decent local business had ten organic slots to compete for and a reasonable chance of a click if its listing looked right. When someone asks an assistant who can fix a burst pipe on a Sunday, the answer is a short list of names with a sentence on each. A ranking with ten slots has room for a business that is merely good. A short list does not.
The symptom shows up in Search Console before the cause is obvious: impressions hold up while clicks do not. Our own site is an example. Over three months it received 45,793 impressions and 82 clicks, and 309 queries containing "best" or "top" produced 7,909 impressions and 0 clicks. Those are exactly the list-style questions an answer now settles on the results page.
Where small business search clicks went
The broadest measurement is SparkToro's. Using Similarweb's US clickstream panel, it found that 68.01% of Google searches in the first four months of 2026 ended without a click, against 60.45% in 2024, and that searches producing a click of any kind fell by 9.51 points (SparkToro). The 2024 figure came from a different panel, so treat the comparison as a direction rather than a precise change.
SparkToro attributes much of that to AI Overviews. Pew Research Center's browsing data from 900 US adults showed people clicked a traditional result in 8% of visits when an AI summary appeared, against 15% when none did (Pew Research Center). Ahrefs found that on December 2025 data, the presence of an AI Overview correlated with a 58% lower average clickthrough rate for the top-ranking page (Ahrefs). AI Mode is not yet the main cause: SparkToro found only 0.34% of searches went to AI Mode from January to April 2026.
The loss is not even across query types. SparkToro notes that some categories still benefit from SEO, including branded searches, local businesses and high-intent transactional queries. In practice, a bakery that got traffic from a post on how to store sourdough has more to lose than one that gets found when people look for a custom cake nearby. Informational content has not become useless; its job has shifted from catching clicks to being the material an engine reads before it decides whom to name. See zero-click for the term.
AI recommendations as a local discovery channel
BrightLocal's 2026 survey of 1,002 US adults is the clearest local evidence (BrightLocal). It found:
- 45% of consumers had used AI tools for local business recommendations, up from 6% in 2025, which makes AI the third most used tool for local recommendations behind Google and Facebook and ahead of Yelp and TripAdvisor.
- Use of Google reviews slipped from 83% to 71%.
- 64% of 30-44 year olds had asked an AI tool for a business recommendation in the past year, against 24% of those over 60.
- ChatGPT was used for business recommendations by 31% in the last 12 months, and Google's AI Mode by 23%.
The age split matters for planning. If most of your customers are over 60, the shift is slower for you, but nearly a quarter of that group already asks assistants for business recommendations, which is not a number to plan around as zero.
Yext's 2026 survey of 3,848 adults who search for local businesses, which Yext describes as self-reported and directional, found 42.7% had used an AI tool for local search in the past month and 28% had tried a new local business because of an AI recommendation in the past six months (Yext).
Small businesses themselves see less of this in their traffic reports. In WordStream's survey of over 300 US small businesses, only 18% count AI among their main traffic sources (WordStream). Both can be true at once: the choosing happens in the chat, and only part of it turns into a visit.
Fewer visits, further along
The visits that do come from assistants look different. In Adobe's data on US retail sites, AI traffic in March 2026 converted 42% better than non-AI traffic, a reversal from March 2025, when it converted 38% worse. Visitors from AI sources spent 48% longer on site and browsed 13% more pages per visit (Adobe). That is large retail sites, not small local businesses, so read it as a direction.
The reason is that much of the comparison happens before the click, and the click is often a check. Yext found only 5% of people move directly from an AI answer to a purchase. After an AI recommendation, 53% search Google or Bing to verify or learn more, 49% visit the business's website directly, 42% click the sources the AI provided, 28% look for reviews and 20% check social media profiles. BrightLocal found 97% of AI users sometimes double-check AI recommendations against real reviews.
Read the figure as a list of places where a recommendation can fail. A business that is named but has an out-of-date phone number on its own site, thin reviews, or a search result that disagrees with the assistant loses the customer at the verification step, after the hard part already went its way.
| Question | Search results era | Recommendation era |
|---|---|---|
| Where the comparison happens | Across several tabs, after the click | Inside the chat, before the click |
| What your website is for | Being found | Being found, and being checked |
| What else gets checked | Mostly your own page | Search results, reviews, cited sources and social profiles |
| Cost of being invisible | Less traffic | Not being considered at all |
| How you find out | Rank tracking and Search Console | Asking assistants a fixed set of questions |
What a small business can control
Nobody outside the engines controls which names appear in an answer, and anyone selling a guaranteed placement is guessing. What you can control is whether you are eligible and whether the facts about you agree with each other.
| Area | Why it matters | What to do this month |
|---|---|---|
| Crawler access | An assistant cannot read a page its crawler is blocked from | Check robots.txt and firewall rules for the search agents |
| Readable pages | Most AI crawlers do not execute JavaScript | Make sure prices, hours and services are in the HTML |
| Fact consistency | People check recommendations against search results, reviews and your site | Align name, address, hours and services everywhere they appear |
| Reviews beyond one platform | BrightLocal's CEO argues AI tools cannot see inside Google's reviews | Collect reviews on the other sites your customers use too |
| Third-party mentions | Claims confirmed elsewhere are easier to repeat | Directories, local press, supplier and partner pages |
| Question-shaped pages | A passage that answers one question completely is easier to lift | One page or passage per real customer question |
On crawler access, the distinction that matters is between training and search. OpenAI separates GPTBot, which crawls content that may be used for training, from OAI-SearchBot, which builds the index behind ChatGPT's search features (OpenAI). Blocking the first is a defensible choice; blocking the second removes you from ChatGPT search results.
On readability, machine-readable pages are not a given even for large retailers: Adobe's checker scored US retail homepages at 75% on average and product pages at 66%. For third-party mentions, our guide to third-party corroboration covers how to find and fix the sources that describe you, and local GEO covers listings and reviews for businesses that serve an area.
How to measure any of this
Referral counts alone will understate what is happening, because so much of the decision is made before anyone clicks. Track three things instead:
- Mentions across a fixed question set. Write the questions a customer would ask without knowing your name, ask several assistants, and record who is named and which pages are cited. Keep the brand name out: in one of our own runs, 13 of 15 generated questions contained it, and the score read 13% when the honest value was zero. Building a prompt set explains how.
- Impressions against clicks in Search Console. Google's Search Generative AI performance reports, announced on June 3, 2026, show impressions in AI Overviews and AI Mode by page, country, device and date, though the announcement lists no click metric (Google Search Central).
- Branded search and direct visits. If people verify recommendations by searching your name or visiting your site, those numbers are where it shows up.
Run the same question list on a fixed schedule and store the answers. A single snapshot of a system that varies from one answer to the next tells a convincing story that may not be true; a run of comparable snapshots shows the direction for your category. Whether that work is worth it for your business is the subject of Is GEO worth it for small businesses?
Sources
- SparkToro, In 2026, less than one third of Google searches still send a click (June 2026).
- Pew Research Center, Google users are less likely to click on links when an AI summary appears (July 2025).
- Ahrefs, AI Overviews reduce clicks, updated study (February 2026).
- BrightLocal, Local Consumer Review Survey 2026: AI and trust (March 2026).
- Yext, Consumer Search Behaviors Report (2026).
- WordStream by LocaliQ, Small Business Website Trends Report (last updated March 2026).
- Adobe, AI traffic surge to retail sites (April 2026).
- OpenAI, Overview of OpenAI crawlers; Google Search Central, Search Generative AI performance reports (June 2026).
Every figure above was checked against the publisher's own page. An earlier version of this article carried more statistics; those we could only find on aggregator sites were removed, and BrightLocal's 83% to 71% is now described as what it measures, use of Google reviews, not Google's share of local discovery.