Something broke in digital marketing over the past two years, and it wasn’t the algorithm. It was the assumption underneath almost every growth plan: that a good ranking eventually turns into a visit, and a visit eventually turns into revenue.
Google still owns search – about 90% of the global market by most counts – but it increasingly answers the question itself, on the results page, before anyone clicks. SparkToro and Datos found that 58.5% of US Google searches ended without a click to the open web. By February 2026, BrightEdge measured AI Overviews on roughly 48% of tracked queries, up from about 30% a year earlier, and at its May 2026 developer keynote Google said those summaries had passed 2.5 billion monthly users. The tap that used to pour traffic is still attached. It just drips now.
I spent the past week reading primary research instead of the roundups, and the founders still growing through search have changed five things. None are glamorous. All of them follow from the same shift: Google stopped being a directory of websites and became an answer machine that cites witnesses. Your job is to be a witness worth citing – and then to own the relationship after the citation.
One caveat before the list. These habits reward patience, and the click data is noisier than most agency decks admit. Where the evidence is thin, I’ve said so rather than rounding it into a confident number.
Habit 1: Answer first, rank second
The clearest evidence on what an AI summary does to behaviour comes from the Pew Research Center. Its July 2025 analysis tracked the real browsing of 900 US adults and found that when an AI summary appeared, users clicked a traditional result on just 8% of visits – nearly half the 15% rate on pages without one. Clicks on the sources cited inside the summary itself? One percent. Sessions were also more likely to end right there, 26% versus 16%.
Presence alone no longer saves you. Seer Interactive’s 2026 study of 53 brands, 5.47 million queries and 2.43 billion impressions found that being cited in an AI Overview earned 120% more organic clicks per impression than being absent from it – yet still trailed comparable queries with no AI Overview by 38%. A citation is a foothold, not a restoration.
So the habit is structural, not stylistic. Put a direct one- or two-sentence answer directly under every question-shaped heading, before the nuance. Give it a number, a name and a date. Comparison tables beat metaphors, because extraction tools lift the table and leave the poetry behind. A reader – or a model – should be able to quote the second sentence of a section and have it stand on its own.
Google disputes the doom. Responding to Pew, the company called the study’s “flawed methodology and skewed queryset” and insisted it “consistently directs billions of clicks to websites daily.” That deserves a fair hearing; Google does send enormous traffic. But a randomised field experiment from researchers at the University of Pennsylvania and Northeastern, published in August 2026, found that hiding AI Overviews increased clicks to websites with no measurable drop in user satisfaction, and that forcing users into AI Mode made them less likely to click out and more likely to try a rival engine. My read: the aggregate can look healthy while the distribution quietly moves against you – and the distribution is all you actually own.
Habit 2: Stop counting sessions, start counting citations
Most dashboards still lead with traffic, and traffic is now the least stable number in the report. AIOSEO’s 2025 analysis found that 94% of web pages get zero traffic from Google and 95% have zero backlinks. Publishing volume was never a strategy, but it is now actively misleading.
The replacement metrics are less familiar and more useful: the share of relevant queries where your page is cited inside the AI answer, query-level click-through split by whether an AI Overview appeared, branded search volume, and the conversion rate of AI-referred visitors tracked separately from organic.
Here is where I part company with the loudest advice. The claim that the average AI visitor is worth 4.4 times a traditional organic one comes from a Semrush projection and gets repeated as settled fact. Baseline Labs’ July 2026 review found the evidence genuinely conflicts – some studies replicate the multiplier, others don’t, and the eye-catching growth numbers (AI traffic to US retail sites up more than 1,300%) start from a base close to zero. Do not import someone else’s multiplier. Segment by referrer in your own analytics and find out.
Where the clicks went, by query type, in one 2026 analysis of 21.7 million Search Console impressions:
Source: Visionary Marketing’s 2026 AI Overview traffic analysis, based on 21.7 million Google Search Console impressions. Figures are year-on-year changes to organic clicks; treat them as a direction, not a forecast for your site.
Habit 3: Make your expertise visible, because Google now checks
Google’s own documentation is unusually blunt about this. In its guidance on creating helpful, reliable, people-first content, it defines E-E-A-T – experience, expertise, authoritativeness, trustworthiness – and says plainly that trust is the most important member of that family. It then asks creators to answer “who, how and why”: who made the content, how it was made, and why it exists. A founder with a real byline, a background page and evidence of their own work answers all three. An anonymous content farm answers none.
Reviews are the same signal in public. BrightLocal’s 2025 survey of 1,026 US adults found 83% read reviews on Google, 71% read them regularly, and 89% expect owners to respond to both positive and negative reviews – not just the flattering ones. The counterintuitive finding: recency matters less than the folklore suggests. Only 20% said a review needed to be within the last two weeks to influence their decision, down from 25% two years earlier.
If I were starting a service business today, I’d spend my first marketing dollar on original evidence and a visible identity, not a content hub. Publish the numbers only you have – pricing, lead times, failure rates, whatever your customers keep asking. That is the material an answer engine cannot synthesise from eleven competitors’ listicles, and it is the material humans screenshot.
Habit 4: Treat “near me” as a channel, not an afterthought
Local intent has been the most durable part of search through every AI change. Around 46% of Google searches carry local intent, and 72% of people use Google when looking for a local business. The reason local is resilient shows up in the table above: local queries lost only about 18% of their clicks to AI Overviews, against 61% for informational queries and 24% for commercial ones. When someone wants a plumber at 7am, they want a plumber, not a two-hundred-word preamble.
That asymmetry is why “near me” deserves its own budget line. It means a fully built-out Google Business Profile, consistent name-address-phone data, a steady trickle of fresh reviews, and suburb- or service-level pages that contain genuine local detail rather than a swapped-out place name. It is unglamorous work, and it is exactly why a lot of owners eventually hand the map pack and the AI answer to someone who specialises in SEO in Sydney rather than a generalist content agency that has never had to rank a service area.
The businesses that win here stop treating local listing work as a one-off setup task and start treating it as a channel with its own cadence, the way they’d treat email or paid.
Habit 5: Rent attention, but own the relationship
Diversification is not a lack of focus; it is an acknowledgement that discovery is now rented. Average Google Search Ads cost per click hit $5.26 in 2025, up about 13% year on year, while SEO is estimated to return roughly 8x against PPC’s 4x over comparable conditions. Email continues to clear around $36 for every $1 spent. The three aren’t competing – they do different jobs at different points in the funnel.
The habit worth copying is this: use paid and social to manufacture demand, then route that demand into something you own – a list, a community, a product habit – before the platform changes its pricing or its algorithm. Branded search is the tell. Branded queries lost only about 3% of their traffic to AI Overviews, the smallest hit of any query type, because people who already know your name search for it on purpose.
This is the one call I’d genuinely ask you to sit with: you can buy reach this quarter and watch the cost creep up every year, or you can spend the same money building an audience that gets cheaper to talk to the longer you keep it. Most entrepreneurs pick the first because it shows up in the dashboard on Friday. I’d argue the second is the only compounding asset on this list.
The uncomfortable part: where this breaks down
None of it is clean. The measurements of how often AI Overviews actually appear disagree by seven percentage points – Similarweb put it at 43% in July 2026, BrightEdge near 48% in February, and Google’s own figure around 50% for that month. Those panels count queries differently, so anyone quoting a single prevalence number as settled is overstating what the data supports.
The conversion-quality evidence is worse. Whether AI-referred visitors are worth more, the same, or less depends on which study you read, and nobody has published a clean randomised test of that yet. If you work in a legally or medically sensitive niche, the trust bar sits higher, and a byline alone won’t clear it. And no dashboard will ever show you the person who read an AI summary, formed a preference, and searched your brand name three weeks later. Some of this success is real and permanently invisible. Measure what you can, and resist the urge to invent the rest.
FAQ
Is SEO still worth doing in 2026?
How do I get my business cited in Google’s AI Overviews?
What should a small business measure if traffic is falling?
Do online reviews still matter now that AI summarises them?
How long does SEO take to work?
The founders getting this right aren’t doing more marketing. They are doing a different kind, aimed at being the source rather than the destination: clear answers, visible expertise, local proof, and an owned audience to catch whatever the answer machine leaves behind. It is cheap to start and slow to pay, which is precisely why it still works.
