/fk
WorkPricingAboutBlogLet's talk

Small Businesses Are About To Get Better Leads From Search - If they do it right

The search behaviour is changing and therefore the opportunities for small businesses. Here are some data points and my personal take on SEO for small businesses.

By Published Aug 26, 202612 min readProofread by AI because I'm ESL
SHARE

My Thesis

Small businesses that have a clearly defined ICP, and a good on- and off-page structure and a decent amount of social signals will see a steady stream of leads.
Flo Krause

How Do I Come Up With This Thesis?

After taking a break from SEO for over a year to focus on other parts of my career, I found myself watching how people search. Not on purpose, it just kept happening. The way search has changed in the last eighteen months is just mind-blowing to me.

What I noticed in myself and in other people is that searching got more detailed. Instead of throwing two keywords in and hoping the result is the right one, people take their time and write down what they are actually looking for. Nobody did that for the last twenty years, because it did not work.

Search is getting more granular, and that increases the chance for a small business to get the last click.

Most of what I read says the opposite. The argument goes that Google now keeps the answer, the click never happens, and the small site starves.

The Search Box Got Longer

For about twenty years people typed two or three words into Google and scanned the results. "plumber berlin." "crm software." That habit is breaking down.(Luckily 🕺)

Search-enabled ChatGPT prompts grew from 4.7 words to 8.7 over twelve months in Semrush's US clickstream panel (Semrush). People stopped translating themselves into keyword language and started describing their situation.

The queries ChatGPT writes for itself got longer too. Peec tracked 20 million of the machine-generated fan-out queries sitting behind those prompts and found their average length roughly doubled, from about six words in October 2025 to twelve by January 2026, with a near-identical trendline across all five countries it sampled (Peec AI).

So specificity is rising from both directions. The person types more, and the machine turns what they typed into longer, sharper searches of its own.

The number I find most useful sits back in the Semrush study. For most of the period they measured, between 65% and 85% of prompts had no match in Semrush's 27-billion-keyword database (Semrush). The way people phrase things to ChatGPT largely does not exist as tracked search volume.

The queries ChatGPT writes for itself doubled in length in four months. A typical Google query has not moved in twenty years. Source: Peec AI, 20 million ChatGPT fan-out queries, Oct 2025 to Jan 2026.

Longer More Detailled Searches

I assumed people were asking a long chain of follow-up questions and closing in on an answer over many turns. Semrush's panel logs about 1.75 queries per ChatGPT session as of February 2026. It sat flat at roughly 1.2 for most of 2025 before jumping in the final four months (Semrush). Semrush reads that jump as power users going deeper. Either way it is nowhere near the back-and-forth I had pictured.

What happens instead is that someone types their whole situation into the first message, and ChatGPT breaks that single prompt into two or three sharper searches on its own. The number of those sub-searches barely moved even as their length doubled, sitting at 2.4 to 2.8 per prompt in October and 2.3 to 2.8 in January (Peec AI). Peec's read is that the machine is making each search more precise rather than running more of them.

The narrowing used to take four searches. Now it happens inside one message, and ChatGPT fans it out. Source: Peec AI, 20 million fan-out queries, Oct 2025 to Jan 2026.

Somebody Will Click And It Will Be The Right Person

There is a fair objection here, which is that a mention inside an AI answer is worth nothing when 68% of US Google searches already end without a click, up from 60% in 2024 (SparkToro). Those two figures come from different panels, so treat the gap as directional rather than precise.

What keeps the click alive is that people are trusting these answers less while using them more. In 2025, 82% of US consumers said AI search was more helpful than traditional search. Asked the same question a year later, 54% did, and outright skeptics went from 3% to 17%. However, 70% said they use AI tools for search more than they did last year (Search Engine Land).

Where that combination ends up: 50% of people who use AI-featured search daily say they click through to a cited source, against 14% of infrequent users (GWI via Search Engine Journal). GWI reads this as habit rather than doubt. These are people who treat the AI answer as a starting point and the cited source as the destination.

It is self-reported, with no published sample size, and it says nothing about how often any single search produces a click. But the direction is the part that matters to a small business. The deeper somebody goes into AI search, the more likely they are to go looking for the source. And assuming that we will see way more power users in the future, that source can be you.

Being Big Stopped Helping

Google ranks you higher when other websites link to you (Don't feel offended SEOs; it's just simplified🫶🏼). Twenty years of SEO were mostly a fight over that, and it was a massive battlefield.

Inside an AI answer it does close to nothing. Surfer analysed five million cited URLs across AI Overviews, AI Mode, ChatGPT and Perplexity and found domain authority barely registered. Correlations with PageRank, harmonic centrality and Surfer's own Domain Score all landed between -0.07 and +0.03, close enough to zero to be noise (Surfer).

So the thing that kept a five-person company off page one for twenty years does not apply to the answer your customer is reading now.

84%

of ChatGPT's citations go to domains outside the top 100

80%

of AI citations, in a 2025 study of long-tail queries, went to pages sitting outside Google's top 100 for that query

2.36%

is all the top-cited domain gets. Reddit is ChatGPT's most-cited source and still takes under three percent

How Social Signals Are Getting More Important Than Ever For Small Businesses

When I say social signals I do not mean likes and shares. I mean the mentions themselves: a YouTube video that names you, a LinkedIn article that quotes you, a Reddit thread that recommends you. Those are the pages AI engines pull from when they build an answer.

Social platforms make up about 5.54% of all AI citations. Within that slice, Reddit accounts for 46.4%, YouTube for 31.8% and LinkedIn for 13%, with Reddit and YouTube together carrying 78.2% (OtterlyAI). A separate report from Wellows found the same rank order across 350,000 citations from five engines (Wellows).

The correlation data points the same way. Across 75,000 brands, YouTube mentions correlated with brand mentions in AI answers at 0.71 to 0.74 depending on the engine, ahead of everything else Ahrefs measured. Branded web mentions came in between 0.66 and 0.71. Link metrics were described only as "very weak," far below every brand signal in the study (Ahrefs).

LinkedIn is where this gets interesting for anyone running a small business. Semrush analysed 89,000 LinkedIn URLs cited by ChatGPT Search, Google AI Mode and Perplexity across 325,000 prompts. On ChatGPT Search and on Google AI Mode, content published by individual members made up 59% of all LinkedIn citations. Company pages only came out ahead on Perplexity. LinkedIn articles accounted for 50 to 66% of cited LinkedIn content, while feed posts managed 15 to 28% (Semrush).

What one person writes gets quoted more than what the company account publishes, on two of the three biggest engines. Small businesses have never been on the right side of a rule like that before.

Social platforms are a small slice of all AI citations, and Reddit and YouTube take most of it. Source: OtterlyAI, 100M+ citation instances across six AI engines, March 2026.

What Does This Mean For Small Businesses And Their SEO Efforts

Off-page SEO now means presence rather than links.

For twenty years, off-page work meant acquiring backlinks, and whoever had budget acquired more of them. The signals that predict whether an engine quotes you are mentions, videos, articles and threads on platforms anyone can post to. Backlinks still feed Google's classic index, so this is an addition to the job rather than a replacement.

On-page changed too, and in a smaller way than people expect. Engines pull passages, not pages. A section that buries its answer three paragraphs down gets skipped even when the answer is good.

What decides which passage gets picked is how closely it matches the question. Of the pages ChatGPT retrieved while building an answer, those whose title overlapped 50% or more with the query were cited 20.1% of the time, against 9.3% for pages under 10% overlap (AirOps). AirOps does not define how it measured that overlap or test whether editing a title changes the outcome, so read it as evidence that specificity gets rewarded rather than as a title-rewriting trick.

A page built around "marketing agency" has close to nothing to match against a twelve-word description of somebody's actual problem. A page that names the exact customer and the constraint they are working under matches on contact.

The uncomfortable part is that this is good news for owner-operators and bad news for the way most agencies sell. Getting cited in a Reddit thread, or having a founder's LinkedIn article quoted, does not require a media budget. It requires having something specific enough to say that somebody repeats it. That is a different constraint from the one small businesses have been losing on, and it is one they can actually clear.

Two things temper it. Reddit is the most-cited domain in AI answers and it is a platform you do not control, where self-promotion gets punished and you often compete against a thread about your own category rather than a competitor. And building brand presence is slower than buying links ever was, which favours whoever started earlier.

If you take one thing from this section: your LinkedIn articles and your YouTube appearances are doing SEO work now, and your backlink budget is doing less of it than it used to.

Here are my practical step for small businesses

01
Your Website Needs To Speak to your ICP
Next to your social presence, your website is the one of your biggest assets in the future. Does your website speak to your ICP in every sentence? Do you outline the pain points? Do you speak their language? Is the website helpful?
02
Make Sure LLM's Can Read Your Website Properly
LLM's preferring websites that have structured data implemented. The reason is simple...To save time and energy, they prefer websites that are easy to read, rather than extracting information. Go to schema.org and implement all necessary markups.
03
Share Your Expertise On Social Media (LinkedIn, Youtube)
Personal branding is more important than ever. By sharing your expertise not just on your website you are widen your net and most importantly you increase your visbility across multiple channels.
04
Make in-house data your biggest asset
Almost everything can be replicated by AI nowadays but your in-house data (Sales-Calls, Customer pain points, success stories etc.) are unique to YOUR brand. Make the most use of it and share it with the world.

Was My Thesis Right?

My thesis was that small businesses that have a clearly defined ICP, and a good on- and off-page structure and a decent amount of social signals will see a steady stream of leads. Based on the data, I would argue it is going in that direction. With one big condition attached, which I will get to.

The part I am confident about is that the barrier came down. Domain authority no longer decides who gets quoted, citations spread across thousands of domains instead of concentrating on page one, and the pages that match a specific question get picked over the pages that match a category. Those are four separate studies pointing the same way.

And something is working against me. The strongest predictor of visibility is now how often people mention you, and being widely talked about is a slower process. Big Brands with bigger budgets and more resources can execute that strategy faster but not necessarily better.

Which brings me to the condition. All of this only holds IF small businesses take the change seriously.

That means rethinking your content strategy and positioning yourself as the expert across the channels that matter. I know this sounds painful, and I know 90% of people do not want to do it. But if you want to stay relevant and credible, you have to put yourself out there and show what you know. Only then do your chances of being visible go up.

I wish you good luck with it. I believe in you 🫶🏼

Frequently Asked Questions

No, but the name is closer to dead than the work is. Getting found still decides whether customers reach you. 68% of US Google searches now end without a click, up from 60% in 2024, and AI answers take a growing share of what is left.
Most people call it AEO or GEO right now, and I do not think either name lasts. SEO in 2015 and AEO in 2026 describe the same job. My bet is that within two to four years one term covers both, and it will not be one of today's acronyms.
SEO aims to win a click from a results page. AEO aims to be quoted inside an answer. The tactics differ for now. In a 2025 study of long-tail queries, only 12% of AI citations also ranked in Google's top ten. The job underneath is the same, which is why I expect the two labels to merge.
For Google yes, for AI answers barely. Ahrefs studied 75,000 brands and found YouTube mentions correlating with brand mentions in AI answers at 0.71 to 0.74, and branded web mentions at 0.66 to 0.71. Link metrics landed below every one of those brand signals. Ahrefs described them only as "very weak."
Inside AI answers it works better than it used to. Domain authority no longer predicts citations, so the barrier that kept small sites off page one does not apply. 84% of ChatGPT's citations go to domains outside the top 100. Pages whose title matches the question get cited about twice as often as pages that barely match.
The query got longer and authority stopped deciding. Search-enabled ChatGPT prompts grew from 4.7 to 8.7 words in a year, so ChatGPT now searches against a description of somebody's situation rather than a keyword. Of the pages it retrieved while building an answer, ChatGPT cited those whose title matched the question 20.1% of the time, against 9.3% for the ones that barely matched.