If you run a small business and your traffic dropped sometime in the last twelve months, you're not alone, and you're not imagining it. Something changed. Several things actually, all at once, and most of the takes online pick one side and run with it.
The honest map is more complicated.
Search didn't die in the age of AI. It also didn't stay the same. It split into a few distinct surfaces that behave differently and reward different things, and the work of being found on the web has shifted around the edges of every one of them. None of that means SEO is over. It does mean that some of the advice from three years ago no longer holds.
This post is the map we'd hand a client who's been swimming in the noise and wants a real picture before deciding what to do. Five real shifts. The genuine upside of each one. The genuine downside. Where the hype is. What hasn't moved. And, at the end, a short opinion on where we'd actually put our hours if it were our money.
The summary box at the top of Google that answers the question directly, drawing from multiple sources. Appears on a meaningful share of searches and changes the rules above the regular blue-link results.
Shift one: Google's AI Overviews and what they did to traffic
Google's AI Overview is the summary box that often appears at the top of a search results page, before the regular results.
The shape of the data, in plain numbers. AI Overviews now appear on something like 26% to 48% of searches, depending on whose methodology you trust. Where they appear, click-through rates drop sharply. Field studies put the hit at 38% to 61%. Zero-click search, the share of queries where the user reads the answer and never visits a website, climbed from 54% to 72% on AIO-triggered queries between 2024 and 2026.
So the easy story is "AI Overviews killed search traffic." It's the half-true one.
The fuller story is two-sided.
On the bad-news side, the hit lands hardest on small businesses. Chartbeat's March 2026 analysis found that small publishers lost roughly 60% of their search referral traffic between 2024 and 2026, while large publishers lost only 22%. The gap is three-to-one, and it's structural. Big sites have diverse traffic sources to cushion the loss. Small ones often don't.
Only about 1 in 100 AI Overview citations actually leads to a click on the cited source. Even when your business appears inside the box, the direct traffic dividend is small.
Small publishers, 2024 to 2026
Chartbeat, March 2026. Large publishers lost 22% over the same period.
On the good-news side, the loss isn't even across all queries. AI Overviews show up most on informational queries (someone asking "how does composting work"), much less on commercial or transactional ones (someone Googling "best dentist Brighton" or "buy ergonomic desk chair"). If your business serves a "ready to buy" search, the AIO hit on your category is smaller than the headline number suggests.
And brands cited inside an AI Overview earn roughly 120% more clicks per impression than uncited brands on the same query. The traffic that does still come through consolidated around fewer, more trusted sources. Being in the box matters more than ranking #2 below it.
There's even a tentative sign the bleeding may be slowing. Organic CTR on AIO queries climbed from a December 2025 floor of 1.3% to 2.4% by February 2026. Could be reversion to the mean, could be noise, could be users learning to look past the box.
Rank for the keyword, get the click. The blue-link economy at full strength.
Get cited inside the AI box, get brand recognition. Fewer clicks. A different game above the same blue links.
The honest summary. Traffic from informational queries is down, often a lot more if you're small. Queries closer to a purchase are largely intact. Being cited matters more than ranking. The share of citations going to small sites is small but climbing slowly. All of that is true at once.
Shift two: a second search surface emerged
People don't just type questions into Google anymore. They type them into ChatGPT, Perplexity, Gemini, and Claude. Those tools became their own search surface, with their own rules.
The underlying AI that powers tools like ChatGPT, Claude, and Gemini. When used to answer questions instead of write text, the same models are often called "answer engines."
The numbers, again with a range. AI-referred traffic to websites grew 527% year-on-year per Search Engine Land's 13-month analysis. ChatGPT alone drives roughly 87% of that referral traffic. Perplexity sits around 12%, Gemini smaller. The total volume is still small in absolute terms compared with Google organic, but the trendline is steep.
Three things make this surface interesting for a small business.
The traffic that does come through converts unusually well. ChatGPT visits convert at roughly 15.9% on average. That's multiples above what most non-branded organic traffic converts at. Visitors arrive having already done their research with the AI. They're closer to a decision than the typical Googler is.
The SERP order doesn't transfer. About 90% of the sources ChatGPT cites originate from beyond the first two pages of Google search results. A site that's never ranked anywhere near page one can still be cited in an AI answer. That's a genuine opening, the first one most small operators have had in years.
And the field is thin. Around 90% of brands currently have zero AI search mentions at all. The early-mover dividend is real, because there isn't a queue yet.
The downsides are honest too.
The volume is small. Even at 527% growth, ChatGPT referrals are still a fraction of what Google organic sends. For now, this surface is a supplement, not a replacement.
It's volatile. AI referral traffic dropped 42.6% from a July 2025 peak before climbing again. The platforms are still figuring out how to cite, when to show links, how to weight sources. A pattern that works today might not work in six months.
And it's harder to measure. You can see that a visitor came from chat.openai.com. You usually can't see which question or which conversation surfaced your page.
This is the most interesting opening in search for small businesses in years. Worth tracking, even when it's not yet a big channel.
Shift three: producing great content at quality just got dramatically easier
This is the most overhyped and underexplored shift of the lot. Worth walking carefully.
Something genuinely changed. A small team, a solo founder, a two-person marketing function, can now do work that used to need a content team. Research, briefing, drafting at consistent quality, structural rewrites, schema markup, internal linking maintenance. All of it cheaper, faster, more accessible than it was 24 months ago.
The good-news side. The capability ceiling for in-house content rose substantially. Things that needed a £4,000-a-month agency in 2022 can be done in-house in an afternoon by a competent owner with a clear brief and an editor who knows what they're looking for.
The output is repeatable. AI tooling lowers the variance of what gets shipped. A solo marketer can produce three or four deeply-researched articles a month without burning out. That cadence used to take a small team.
Production isn't the bottleneck anymore. The bottleneck is editorial taste. Knowing what to write, what makes it good, what to cut. That's a much more interesting bottleneck than "we can't afford a writer."
So this is the real equaliser of the shift. Small businesses have closed a meaningful gap on bigger competitors when it comes to in-house content production.
Now the bad-news side, because there's a trap inside the new capability.
Everyone has the same tools. Your competitors do, the next entrant does, the one-person shop you've never heard of does. The advantage decays unless you're using the new capability on something genuinely differentiated.
Volume is worse than nothing. Google's March 2026 core update specifically penalised sites publishing 1,000 or more unedited AI articles, with traffic drops of 40% to 90%. Sites with smaller volumes of human-edited AI content saw 30% to 80% gains. The lesson is direct. Volume without editorial discipline is now an active liability.
The volume trap
Google's March 2026 update penalised sites publishing 1,000+ unedited AI articles with 40-90% traffic drops. Sites with smaller volumes of human-edited AI content saw 30-80% gains. Volume isn't the moat anymore. Editorial discipline is.
The bar for "good enough" rose. Generic listicle-style content that ranked in 2023 doesn't rank in 2026. Differentiation has to come from real expertise, genuine experience, original framing, or first-party data.
The interesting bit. Ahrefs studied 600,000 top-ranking pages and found that 86.5% contain some AI-generated content. The correlation between AI-content percentage and ranking position is 0.011, statistically zero. Google isn't trying to detect AI. It's trying to detect thoughtlessness. AI is just a faster way to be thoughtless if you let it.
So the gap really has closed, but only if you bring an editor to the work.
Shift four: competition got fiercer, more crowded, and stranger
The flip side of the production shift. Because shipping a competent-looking article got easier, more sites publish more content faster. The total volume on any given topic went up. The marginal effort required to ship a "good enough" article went down. Combine those, and the field gets noisy.
The encouraging side first.
There's still a quality ceiling, and most of the new entrants don't reach it. Mass AI output is mostly mediocre. A small business that takes content seriously can stand out, arguably more easily than before, because the average is dragged down by the volume players.
The reward for genuine expertise widened. When everyone can produce plausible text, the verifiable signal of actual experience becomes worth more. Named authors. Real photos. Original observations. First-party data. Anything that proves a real person with a real opinion sat down and wrote the thing. E-E-A-T isn't a Google catchphrase anymore. It's how AI tools decide whom to trust.
Long-tail and niche content became more viable. Writing that wouldn't have been worth a content team's time becomes worth a solo founder's afternoon. Small operators can compete on coverage breadth in a way they couldn't before.
The harder side.
Established brands compound faster. They have existing authority signals (links, mentions, branded search volume, an audience) that the AI tools weight heavily. The gap between "no brand" and "some brand" matters more than it used to.
Discovery is harder. With more content competing for fewer organic clicks, even genuinely good work struggles to find an audience without distribution. Publishing alone isn't enough anymore.
The arms race is real. Tools improve. Tactics commoditise quickly. What worked six months ago is often table stakes today.
The honest summary. The small business's production gains are real, and the field also got harder to stand out in. Both are true at once. Whether you net out ahead depends on what you do with the new capability, which is the conversation worth having.
Shift five: authority moved off-site
This is the shift most generic SEO content gets wrong. Worth surfacing properly.
AI tools weight third-party signals about your business much more heavily than your own site's signals. They need to verify that you are who you say you are, and the way they verify is by checking what other places on the web say about you.
The data behind that claim, in plain numbers.
Brand mentions are now roughly 3x more predictive of AI visibility than backlinks. Backlinks still matter. They just don't matter as much as having other people talk about you.
About 48% of AI citations come from community platforms, Reddit and YouTube in particular. Roughly 85% of brand mentions overall live on third-party pages rather than on your own domain.
Domains with extensive brand mentions on Quora and Reddit have around 4x higher chances of being cited in AI answers than those with minimal third-party activity.
"Brand mentions are roughly 3x more predictive of AI visibility than backlinks.
SourceAirOps, 2026 State of AI Search
The opportunity side.
It's not a closed ecosystem. You can earn off-site mentions actively. Get quoted in industry publications. Appear on podcasts. Answer real questions on Reddit and Quora in the subreddits and topics your buyers actually use. Get listed in genuine directories. Build a YouTube presence in your niche. Most of these routes are accessible to a small business with patience.
There's a clear early-mover dividend. The 90%-of-brands-have-zero-AI-mentions figure is a real opening for businesses that move now. The competitive field is thin and the work compounds.
The signal is harder for competitors to fake. A few hundred genuine mentions across credible sources is more durable than the link-building tactics of the 2010s. Reputation work is slow, but it sticks.
The downsides.
You can't fully control the signal. What other sites say about you is largely outside your hands. Reputation work is patient work.
It rewards visibility budget. Time spent appearing in podcasts, getting quoted, being part of conversations. That's investment a small business might not be able to spare on top of everything else it does.
The off-site signals AI tools care about are themselves evolving. Reddit's share of ChatGPT citations dropped from over 14% to around 2% between early and late September 2025 when ChatGPT changed how it weighted that source. The platforms aren't stable yet.
Still, this is probably the shift with the largest gap between "important" and "what most small businesses are doing." Worth taking seriously now, while almost nobody else is.
What hasn't actually changed
Plenty of things didn't move, and it's worth saying so. The reader doesn't need to relearn everything.
The fundamentals of SEO are still the fundamentals. Schema markup, mobile experience, page speed, Core Web Vitals, internal linking, on-page hygiene. None of that disappeared. Most of it matters slightly more than it did, because both Google and AI tools now use the same structured signals to decide whom to trust.
- Google Business Profile. Still the highest-impact local-SEO lever. If your business has any local component, the GBP is still where you start.
- Reviews. 87% of consumers read reviews before contacting a local business (BrightLocal, 2026). Both a ranking signal and the closing pitch the prospect reads after finding you.
- Mobile and page speed. Still hygiene. They don't win SEO on their own, but bad scores actively cost you.
- Schema markup. Still the cheapest high-impact technical work. Now even more important, because schema is how AI tools verify entities.
- E-E-A-T. Experience, expertise, authoritativeness, trustworthiness. Arguably more important than it has ever been. The way both Google and AI tools decide whom to cite.
If you've been keeping the fundamentals tidy, most of that work still counts. The job isn't to throw it out. The job is to layer the new shifts on top of it.
What's hype, and what to ignore
A short, opinionated section. The permission slip to skip a few things you might have been told you must do.
- llms.txt. The proposed file standard for telling AI crawlers about your site. Adoption is around 10%, no major AI provider has publicly committed to reading it, and 8 of 9 sites in OtterlyAI's experiment saw no measurable traffic change after implementing it. Worth a glance in 2027 if adoption shifts. Today, ignore.
- 'AI content humanizers' and detection-evasion tools. Google isn't trying to detect AI. It's trying to detect thoughtless content. The tool you need is an editor, not a humanizer.
- The acronym soup. GEO, AEO, AIO, GAIO, SGE, and whatever launches next month. They mostly describe the same shift from slightly different angles. Pick one mental model. The labels will sort themselves out.
- Re-platforming because of AI. If your site is genuinely slow or unindexable, fix it. If it's fine and you're tempted to rebuild because somebody on LinkedIn told you AI requires a new stack, you're being sold to.
Saving four to six hours a month by not chasing these is worth more than chasing them ever was.
Where we'd put our hours, if we were starting from scratch
The one prescriptive moment in a piece that mostly resisted being prescriptive.
If we were rebuilding a small business's organic strategy from a blank page in 2026, we'd do four things and not much else.
Pick one topical area you know more about than your competitors do. Go deep on it. Eight to twelve genuinely good pieces of writing on a single topic beat a hundred mediocre posts on everything. Topical depth is one of the few signals both Google and AI tools weight heavily, and it's something a small business can compound on with patience.
Spend real time getting mentioned in third-party places. Podcasts. Industry publications. Real Reddit answers in subreddits you care about. Genuine directories. Slow work, durable signal. Most competitors won't bother, which is the opening.
Make sure every piece of content has a named human attached to it. Author bio, real photo, real expertise. The cheapest E-E-A-T upgrade you can make, and the one most generic AI-drafted content doesn't bother with.
Set a fifteen-minute monthly habit. Search your top five customer questions in ChatGPT, Perplexity, and Google AI Overviews. Note where you appear. Note where you don't. That's your visibility scoreboard, and it tells you what to work on next.
If you want to go deeper, our killer SEO content guide covers the writing side of all of the above, and the on-page SEO factors post covers the technical layer in detail. For local businesses specifically, the local SEO playbook is the companion piece to this one.
The real shift in SEO isn't AI versus search. It's that being a credible source on the web, verifiable, attributed, and distributed beyond your own domain, matters more now than it has in twenty years. The small businesses that take that seriously have a window before bigger competitors close it.
Common questions
Is SEO dead in 2026?
No. SEO isn't dead. The mechanics changed. Google's AI Overviews now appear on a meaningful share of searches and reduce click-through rates by 38% to 61% where they appear, with the impact concentrated on informational queries. Commercial and transactional queries are largely intact.
The wider shift is from 'ranking pages' to 'being a cited entity.' Both Google and AI answer engines now look for credible, verifiable sources to reference. SEO didn't die. It got more interesting.
Why did my organic traffic drop in 2025?
If your impressions are stable but clicks are down, AI Overviews are most likely the cause. Open Google Search Console and compare the two metrics over the last twelve months. A stable impressions line with a falling clicks line is the classic AI Overview signature.
If both impressions and clicks fell together, something else changed. Possibly a Google algorithm update, possibly a technical issue, possibly a competitor moved on your space. Cross-reference your data against rankings in a tool like Ahrefs or Semrush to isolate which one.
How do I show up in ChatGPT search results?
Three things matter. Be a recognisable entity (a clear business name, consistent NAP, structured data on your site). Be referenced by third parties (Reddit answers, Quora threads, industry publications, podcasts). And write content with the answer in the first 200 words, since around 44% of LLM citations come from the opening 30% of a page's text.
It's slower and less direct than ranking on Google, but the early-mover field is thin.
What is Generative Engine Optimisation (GEO)?
GEO is the practice of structuring content so that AI answer engines (ChatGPT, Perplexity, Gemini, Google AI Overviews) cite or quote it in their generated answers. It overlaps heavily with classic SEO. Some things get extra weight in a GEO context, including clear declarative writing, structured data, primary-source citations, and being mentioned by other sites the AI tool already trusts.
The acronym soup is loud right now. Don't get lost in it. The underlying shift is what matters.
Should small businesses still invest in SEO?
Yes, with a different emphasis. The classic ranking work matters less than it used to on informational queries. Local SEO, commercial-intent queries, and being cited as a source all matter more. If your customers actively search for what you sell, organic still pays back, just at a different shape than 2022.
Where it stops paying is on broad informational topics that are hit hardest by AI Overviews. That's where the channel has genuinely got harder for everyone.
