The ROI of AI-Search Visibility for Small Business

This question sits on Vector 11, Measure, of Formative Digital's 12-Vector method, the discipline of attaching honest numbers to AI-search work. Most return-on-investment content for generative engine optimization skips that discipline. It runs the same vendor-friendly formula, AI-attributed traffic times conversion rate times lifetime value divided by spend, then headlines a multiplier that does not survive contact with a real small business. We are going to invert the order. Before any upside, an owner should size the prize against two sobering facts: most Canadian small businesses are fighting to survive past year five, and AI engines still send almost no traffic. Only after the prize is sized honestly does citation share, the one input you can actually move, start to look like a sound bet.

ROI for AI search is lead value, not traffic volume

Return on investment for AI-search visibility is the dollar value of the leads your citations earn, set against the cost of earning them, and nothing about that sentence mentions traffic volume. That framing matters because the volume is tiny. Jason Tabeling's 13-month analysis in Search Engine Land, published February 25, 2026, found large language model referral traffic averaging under 2% of total referral traffic, with a per-engine range of 0.15% to 1.5% across ChatGPT, Perplexity, Gemini, and Claude. If you size the prize as raw clicks, AI search looks like a rounding error and you stop reading. The prize is not clicks. It is the small number of high-intent decisions those clicks represent.

There is a credibility reason to start sober rather than excited. Statistics Canada and ISED report 1.08 million small businesses in Canada as of December 2024, with five-year survival at 67.5% in services-producing sectors, so roughly a third of services firms do not reach year five. An owner in that position cannot act on the 300% to 500% return claims or the 162.6x multiplier that GEO vendors reach for; those numbers fail the test any cautious buyer applies. A figure you cannot bank is not ROI. It is marketing.

So the working definition is narrow on purpose. AI-search ROI is a thin stream of qualified leads, valued at what a customer is actually worth to you over their lifetime, earned over months, costed against the work, and reported with ranges instead of a single confident multiplier. Everything below builds that estimate one honest input at a time.

An AI-referred lead is worth more, and far scarcer

An AI-referred lead is worth more than an organic one because it arrives later in the decision and pre-qualified by the engine. Semrush studied the impact of AI search on traffic across more than 500 high-value topics and, in findings released June 9, 2025, reported that based on conversion rate the average AI-search visitor is 4.4x as valuable as the average visit from traditional organic search. Read that as an average across verticals, not a promise for your category. The mechanism is intuitive. When ChatGPT or Perplexity names your business in an answer, the person reading it has already described their problem in a full sentence and received a short, vetted shortlist. They are closer to buying than someone who typed two words into a search box.

Tabeling's dataset reinforces the pattern from the conversion side. In his 13 months of data, LLM referrals were the highest-converting source measured, near 18%, while arriving at roughly 25x lower volume than SEO or direct traffic. That is the entire economic shape of AI search for a small business in one sentence: low volume, high intent. You are not buying a flood. You are buying the few sentences where a buyer asks an engine to choose, and being one of the names it returns.

This is why the right unit of value is customer lifetime value, not a single transaction. If a Brantford HVAC company earns one new maintenance customer from an AI citation, the value is not the first service call. It is the multi-year relationship. Pricing the lead at lifetime value is what lets a small, scarce stream of AI-referred leads still clear the cost of the work. It also keeps the math honest, because lifetime value is a number you can estimate from your own books rather than borrow from a vendor's slide.

Citation share is the one input you can actually control

Citation share is the percentage of relevant AI answers that name your business, and it is the leading input you can move, which is exactly why it belongs at the centre of an ROI case instead of raw traffic. You cannot directly control how many people ask an engine about your category, and you cannot control the click-through that follows. You can influence whether the engine cites you when someone does ask. That influence is not folklore. The peer-reviewed grounding for it is Aggarwal et al., whose paper GEO: Generative Engine Optimization (arXiv:2311.09735) ran a controlled evaluation across the 10,000-query GEO-bench and found that generative engine optimization methods can lift source visibility in AI answers by up to 40% on the Position-Adjusted Word Count metric, with the largest gains from adding statistics, credible quotations, and citations to reliable sources.

Why front-loaded, well-structured content earns that share has its own evidence. Kevin Indig's early-2026 analysis in Growth Memo, examining 1.2 million search results, found that about 44.2% of ChatGPT citations come from the first 30% of a page's text, the pattern he calls the ski ramp. An engine reaches for the dense, answerable material near the top. So the controllable input has a controllable technique behind it: state the answer early, support it with named sources, and structure it for extraction. That is engineering, not magic ranking dust.

What counts as good citation share is local and category-specific, and here Formative Digital's own first-party data sets a realistic bar. Our May 2026 analysis of 1,732 AI-engine citations across nine Ontario cities found that only 16.3% of cited sources were named by two or more of the four engines; 83.7% were unique to a single engine. Being cited by two of the four for your core prompts already puts you ahead of most. The same scrape carried a harder lesson for the small-business owner specifically.

That finding, drawn directly from the scrape, is the difference between this analysis and the vendor version. Directories such as downtownhamilton.org, the city economic-development sites, Yelp, and canadianchoiceaward.ca won the local best-business citations far more often than any individual small-business site. So the highest-return work is frequently off your own domain: earning a place in the directories and review sources the engines already pull from. If you want the operational sequence for that, our guide to how to earn a place in the listings AI engines pull from lays it out by engine.

The honest time horizon runs in months, not weeks

The time horizon for AI-search ROI runs across several months, because the inputs that move citation share move slowly and the traffic base is small enough that growth compounds rather than spikes. Tabeling's 13-month dataset showed LLM referral traffic growing roughly 80% from the first half to the second half of 2025, with a wide company-by-company range of 10% to 300%. An 80% gain sounds dramatic until you remember the starting point is under 2% of referrals. Growth off a small base is still small in absolute terms for a while, which is precisely why a responsible projection is patient.

Three things gate the horizon. Engines re-crawl and re-ground on their own schedules, so a content or schema change does not surface instantly. Directory and review signals, the sources that actually carry local citations, update on their own cadence, often slower than your site. And citation share itself is noisy month to month, because AI answers vary between runs of the same prompt. None of that argues against the work. It argues for measuring across a quarter or more, not declaring victory or defeat after three weeks. If you are weighing this against conventional search, the shift the two timelines reflect is the subject of our look at why AI citations are becoming the new backlinks.

Most AI influence never appears as a tracked click

Most of AI search's influence never shows up as a tracked referral, which is the central honesty problem in any ROI claim and the reason every projection here carries a range. When ChatGPT names your business in an answer and the reader does not click the citation but later searches your name directly, or simply calls you, the credit lands on direct or branded search, not on AI. The influence was real; the attribution was lost. That broken chain is the gap a tidy 300% return claim quietly papers over.

The disciplined response is to measure the leading indicator you can see and stop pretending the trailing one is fully attributable. You can observe citation share by running your priority prompts through the four engines and logging whether you are named. You can watch on-site behaviour from AI referrals in Microsoft Clarity or your analytics, where the small but high-converting AI sessions are visible. What you cannot do is draw a solid line from every AI mention to a closed sale, so you estimate that bridge with a conservative range and label it as an estimate. For the mechanics of capturing what is trackable, our walk-through of the 1,732-citation Ontario dataset covers the method we run on our own scrape. If you would rather have us baseline your prompts and report the gap, a no-obligation citation-share baseline is what we produce, yours to keep either way.

There is also a discipline of what to stop measuring. Raw AI-referral sessions as a headline number mislead, because the volume is structurally tiny and a bad month looks like collapse when it is just variance. Impressions and a rising count of prompts where you appear at all, without checking whether those prompts have buying intent, flatter the report without moving revenue. The metric that matters is citation share on the handful of prompts a paying customer would actually type, valued at lifetime value. Anchoring on that one outcome is also the honest answer to the click that never happens, which is exactly the dynamic our piece on zero-click search for local business works through.

An honest projection for a Brantford small business

An honest AI-search ROI projection for a Brantford small business reads as a range built from your own numbers, not a multiplier borrowed from a vendor deck. Take a local service firm with a customer lifetime value of 1,500 dollars and a realistic ambition: over six to nine months of work, lift citation share enough to earn somewhere between two and six AI-attributable customers. At lifetime value, that is 3,000 to 9,000 dollars of customer value. Whether that clears the cost of the work depends on the engagement, your margins, and how contested your category's citation sources are. State it as a range, show the inputs, and let the owner judge it.

Notice what the projection does not do. It does not promise a 4.4x return; the 4.4x figure is per-visit value relative to organic, not a portfolio multiplier, and treating it as the latter is the exact sleight of hand that fails a careful reader. It does not assume clean attribution; the customer count is deliberately small and framed as attributable, not guaranteed. And it leans on lifetime value rather than a first sale, because that is the only way a scarce, high-intent stream pencils out for a small business. Conservative inputs are not pessimism. They are what makes the number trustworthy.

This is also where Formative Digital's Results Guarantee fits a YMYL-disciplined case, by promising only what is measurable. If an existing domain shows no measurable organic search results after twelve months of work with us, we continue working at no additional cost until results land. It is a continuation of work, not a refund, and it deliberately attaches to measurable search outcomes rather than a promised revenue figure, because revenue depends on factors outside any agency's control. For the same reason, a single proof point such as the Mattress Miracle results in Brantford retail, where organic visits grew from roughly 1,000 to more than 82,400 a month, is shown with the standing caveat that outcomes depend on industry, competition, and existing digital presence. One client's result is evidence, not a forecast for yours.

The honest bottom line: value AI-search visibility as citation share converting into scarce, high-intent leads priced at lifetime value, expect returns over months, and trust only the numbers you can measure. Sized that way, generative engine optimization is a sound bet for many Ontario small businesses. Sized the vendor way, it is a promise waiting to be broken.