Short answer: for most small businesses whose customers ask questions before they buy, GEO is worth it in its cheap form: making sure assistants can read your site, that your homepage and service pages answer real questions plainly, that your listings and reviews say the same thing your site does, and checking once a month whether you are being named. Almost all of that is work you would do for search anyway. It is not worth a separate budget line, a retainer that promises placements, or doing before the basics are in place. Nobody outside the engines can promise you a mention or a timeline.
The rest of this article sets out the evidence behind that answer, what the work costs next to what you already do, when GEO is honestly not worth it yet, and a starting plan a small team can run without new software.
What has actually changed for small businesses
The clearest evidence is about local discovery. In BrightLocal's 2026 survey of 1,002 US adults, the share of consumers using AI tools for local business recommendations climbed from 6% in 2025 to 45%, which makes AI the third most used tool for local recommendations behind Google and Facebook and ahead of Yelp and TripAdvisor. Adoption is uneven by age: 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.
Yext's 2026 survey of 3,848 adults who search for local businesses points the same way, with the caveat that Yext itself calls the data self-reported and directional. 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.
The other side of the picture matters just as much. When small businesses themselves are asked where their website traffic comes from, AI is the source they pick least often, well below what customers report:
In WordStream's survey of over 300 US small businesses, 18% count AI among their main traffic sources, and among one-person shops it is 11%. So the honest summary is that assistants are becoming a place where customers choose businesses, while the clicks that arrive directly from them are still a small slice of traffic. If you judge GEO by referral sessions alone, you will undervalue it; if you judge it by headlines, you will overvalue it.
Search itself is also sending fewer clicks per query. SparkToro's 2026 study found that 68.01% of US Google searches in the first four months of the year ended without a click, against 60.45% in 2024, though the two years were measured on different panels. Being the name inside the answer matters more when fewer people scroll past it.
Is GEO worth it if you do not rank on page one?
Search rankings reward accumulated authority. Being cited in an answer is related, but it is not the same contest, and the gap between the two is where a small business has room.
In Ahrefs' analysis of 863K keyword results pages, 37.9% of URLs cited in Google's AI Overviews also appeared within the first 10 result blocks, 31.2% came from positions 11 to 100 and 31.0% from beyond the top 100. Ahrefs compares this with roughly 76% in its July 2025 study, notes that it also improved its parsing method in between, and suggests AI Overviews now lean more on sources found by query fan-out than on the original results page. That does not mean a weak page gets cited. It means a page that clearly answers a specific question has a chance even when it does not rank on page one.
Specificity is where small businesses often have an edge. A national chain rarely has a page that explains, in plain language, how long a particular repair takes or which material suits a particular climate. A small business answers those questions every week by phone and email. Writing the answer down where a crawler can read it takes writing time, not a marketing department.
The academic evidence is encouraging but should be read carefully. The original GEO paper reports that its methods boosted visibility by up to 40% in generative engine responses on a benchmark of 10000 queries; citing sources, adding quotations and adding statistics worked best, keyword stuffing offered little to no improvement, and results varied by domain. That is a lab benchmark, not a promise for your shop.
There is an honest counterweight. When assistants name businesses for a category question, it is easier for them to name one whose facts are confirmed in places it does not control: reviews, directories, local press. Small businesses usually have fewer of those, which is why the listings and reviews step below is not optional.
What GEO costs, compared with what you already do
An earlier version of this article showed a chart of cost per lead by channel. We removed it because the figures were illustrative, not measured. What we can say without inventing numbers is where the work overlaps with work you already do:
| Step | What it costs | Also helps ordinary search? |
|---|---|---|
| Check that AI search crawlers are not blocked | One check of robots.txt and any firewall rules | No, but it is a precondition for everything else |
| Make key facts readable without JavaScript | Depends on your platform; often nothing | Yes |
| Rewrite the homepage and service pages to answer questions directly | Writing time | Yes |
| Make listings and reviews consistent with your site | Admin time, ongoing | Yes, especially locally |
| Check monthly whether assistants name you | A recurring check by hand, or a tool | No; this is the new part |
That overlap is what small businesses report doing. In WordStream's survey, the most common actions to improve AI visibility were structuring pages with clear headings (35%), improving readability (26%) and technical work such as speed and mobile (24%), and the report notes that just about every tactic is also SEO best practice. Only one row in the table is genuinely new work, and it is the cheapest: asking the same questions every month and writing down who gets named.
The comparison that matters is not GEO against SEO. It is the cheap version of GEO against the expensive one:
| The cheap version | The expensive version | |
|---|---|---|
| What you pay for | Your own writing and admin time, plus a monthly check | A retainer, a separate content stream, a dashboard |
| What it promises | Readable pages and an honest measurement | Often placements or a timeline, which nobody outside the engines can deliver |
| What you own afterwards | Better pages that also serve search and customers | Reports, and whatever content was produced |
| Right for | Almost every small business with a website | Businesses where one more AI recommendation is worth a lot and the basics are already done |
When GEO is not worth it yet
- You have no website and no plan for one. That describes many small businesses: in WordStream's survey, 43% of businesses without a site never plan to launch one. Your priority is accurate profiles you already control, such as your Google Business Profile, directory listings and social pages. Our guide to local GEO covers that case.
- Your pages cannot be read. Most AI crawlers do not execute JavaScript, so a site whose text appears only after client-side rendering is an empty page to them. Fix that before writing anything new.
- Someone is selling you guaranteed placements. Assistants decide answer by answer. A provider can improve your eligibility and show you honest measurements; it cannot promise you will be named.
- Customers do not research before buying. If people buy from you on impulse or by walking past, there are few questions for an assistant to answer, and your time is better spent elsewhere.
A starting plan for a small team
- Check crawler access. OpenAI separates GPTBot, which crawls content that may be used for training, from OAI-SearchBot, which builds the index behind ChatGPT's search features. Blocking the training crawler is a legitimate choice; blocking the search one removes you from ChatGPT search results. Our AI crawler access guide lists the agents for each assistant.
- Confirm your key facts are in the HTML. A quick test is below the list.
- Rewrite the homepage and each service page. Businesses that track AI referrals named the homepage (57%) and product or service pages (48%) as the pages those visits land on most often. Put what you do, for whom, where and at what price level in the first lines.
- Answer the ten questions customers ask most. Take them from emails, calls and reviews, and answer each in a passage that makes sense on its own. Writing for extraction shows how.
- Make your listings and reviews agree with your site. People check what an assistant tells them, and they check in several places.
- Measure monthly with questions that do not contain your name. See the last section.
curl -s -A "OAI-SearchBot" https://example.com/services/ | grep -i "opening hours"
If the text you care about is missing from that output, the page needs server-side or static rendering, not more content.
Step five deserves the evidence behind it, because it is the step small businesses skip most often. An AI recommendation is rarely the end of the customer's research:
Only 5% of Yext's respondents said they move straight from an AI answer to a purchase. Most go on to search, visit your site, open the cited sources or read reviews, and BrightLocal found that 97% of people who use AI for local recommendations sometimes double-check them against real reviews. If your hours, prices or services differ between your site and your listings, this is the moment the customer notices.
How to tell whether GEO is working
Write a fixed list of questions a customer would ask without knowing your name, such as "who repairs sash windows near me" rather than "is our business any good". Ask several assistants, record which businesses are named and which pages are cited, and repeat with the same list each month. Treat a change as real only when it repeats in the next run.
Keeping the brand name out is the rule that matters most. In one of our own runs, 13 of 15 generated questions contained the brand name, and the score read 13% when the honest value was zero. Building a prompt set explains how to write a list that avoids this.
Expect referral numbers to stay small for a while, and do not read too much into them. Visits from assistants are a different kind of visit: in Adobe's US retail data for March 2026, AI traffic converted 42% better than non-AI traffic. That is large retail sites, not small local businesses, so treat it as a direction rather than a forecast. If you would rather not run every check by hand, some of them are available on our free tools site.
Sources
- BrightLocal: Local Consumer Review Survey 2026, AI and trust
- Yext: 2026 Consumer Search Behaviors Report
- WordStream by LocaliQ: Small Business Website Trends Report and LocaliQ: 2026 SEO insights
- SparkToro: 2026 zero-click study
- Ahrefs: AI Overview citations and the top 10
- Aggarwal et al.: GEO, Generative Engine Optimization
- OpenAI: overview of OpenAI crawlers
- Adobe Digital Insights: AI traffic to US retail sites, April 2026