AI SEO Results: What 2026 Studies Show

AI SEO results show lost clicks, not guaranteed replacement traffic

For a Las Vegas law firm, measured AI SEO results support testing AI visibility, but do not establish what a citation is worth compared with a Google ranking. The February 2026 Ahrefs AI Overviews update, using desktop Search Console data across 300,000 keywords, found a 58% lower average clickthrough rate for the top result when an AI Overview appeared. New referral evidence published in September 2026 shows AI traffic growing, while remaining a low single-digit share of most sites' traffic.

That distinction matters when an agency sells visibility as a replacement for search traffic. A mention, a linked citation, a visit and a qualified inquiry are different outcomes. None of the supplied studies establishes a universal exchange rate between an AI citation and a Google ranking.

The evidence below was assembled through September 2026. Pew's July 2025 research remains its latest measurement located in that source review, rather than a new study conducted this year.

The studies measure different parts of the customer journey

Read the measurement column before comparing the effects. A search visit, an AI conversation and a website referral have different denominators. The local implications are our interpretation of the supplied findings, not additional measured outcomes.

Study and publisher	Publication date	Sample or method	What was measured	Measured effect	Meaning for a local service business
AI Overviews update, Ahrefs	February 4, 2026	300,000 keywords, equally split by Overview presence; aggregated desktop Search Console data, December 2023 versus December 2025	Top result clickthrough	58% lower average CTR with an Overview	Ranking does not preserve clicks
Google AI summary browsing study, Pew Research Center	July 22, 2025	900 US adults; 68,879 unique March 2025 searches; results scraped April 7 to 17	Clicks during search visits	Traditional result clicks: 8% with summaries versus 15% without; summary link clicks: 1%	Citation exposure rarely became a click
Generative AI landscape statistics, Similarweb	July 29, 2026	Worldwide traffic panel, June 2025 to May 2026	Platform use and citation prevalence	Monthly AI website visits grew 70% to 9.5 billion; ChatGPT citation share reached about 6.8%	Platform growth exceeds citation opportunity
AI referral traffic by industry, Similarweb	September 3, 2026	Desktop and mobile web across ChatGPT, Gemini, Claude, Perplexity, Grok and DeepSeek; June 2025 to May 2026	Outbound referral visits	770.7 million monthly referrals, up 117.4%; low single-digit traffic share for most sites	Fast growth from a small share
Expanded AI Visibility Index, Semrush	June 26, 2026	126 million US prompts, January through April, across ChatGPT, Gemini, Google AI Mode and AI Overviews	Citations and brand visibility	Average sources per response: ChatGPT 15, Gemini 3; Gemini mention and citation overlap as low as 30%	A mention need not link to you
State of Local SEO, Sterling Sky	June 26, 2026	Agency tracking: 179 Business Profiles across 34 law firms via Jepto; rankings across 322 markets via Places Scout	Local pack inclusion and agency client referrals	AI packs included 5,943 unique businesses versus 18,330 in regular packs; second-year ChatGPT traffic equaled 2% of Google traffic	Fewer local slots, limited replacement traffic

The table supports a practical purchasing rule: ask which outcome the proposed work will improve and how that outcome will be recorded. A vendor reporting citations has not yet demonstrated additional customers.

Lower Google clickthrough does not tell you what AI will return

The February 2026 Ahrefs update compared a forecasted top-position CTR without AI Overviews of 0.037 with an actual December 2025 CTR of 0.016 when an Overview appeared. Those decimals equal 3.7% and 1.6%. The reported headline reduction comes from the study's analysis; rounded figures should not be treated as exact inputs for reproducing it.

For our own worked example, assume a page receives 10,000 eligible impressions. Applying those rounded rates gives 10,000 × 0.037 = 370 clicks and 10,000 × 0.016 = 160 clicks, a difference of 210. This illustrates exposure to lower clickthrough, not a forecast for a particular dealership or law firm.

The July 2025 Pew study measures something else: whether a tracked search visit produced a click. Its 8% versus 15% comparison is a 7 percentage point difference, calculated as 15 minus 8. It does not measure the CTR of the top-ranking page, so averaging its result with Ahrefs would produce a meaningless benchmark.

Pew also found that users ended their browsing session on 26% of pages with summaries versus 16% with traditional results only. Ending a session does not prove the searcher was satisfied, became a customer, or abandoned the purchase. It tells an owner that the next observable website visit may never happen.

A skeptical owner might ask whether appearing inside the answer recovers those lost clicks. Pew's measured summary-link click rate gives little reason to assume full recovery. Exposure could still influence a later branded search or call, but these extracts do not quantify that value.

Platform growth and local opportunity are not interchangeable

Similarweb's July 2026 roundup found ChatGPT's share of generative AI website visits fell from about 76% in June 2025 to roughly 53% in May 2026. Gemini rose from under 9% to around 27% to 28%. A report monitoring only ChatGPT can therefore miss meaningful changes elsewhere, even while accurately recording that platform.

The same roundup reports a change in where visits land. After ChatGPT added homepage links in May 2026, homepage referrals rose from roughly 26% to 29% of its referral traffic to about 62% to 63% by late May. An increase in homepage visits can reflect product design as well as better content or stronger brand visibility.

Industry differences also matter. In Similarweb's September 2026 report, 22.6% of Travel answers included web citations, against an all-conversation average of 6.8%. The July roundup put Professional Services below 4%. A hospitality operator and a law firm should not accept the same citation expectations merely because both serve Las Vegas customers.

Semrush's June 2026 index adds another constraint: mentions and citations can diverge. If an answer names your firm but links to a directory, the directory receives the immediately available click. Reporting both events as equivalent visibility hides a practical difference for the owner paying the bill.

Our AI SEO work in Las Vegas should be evaluated against that same distinction. The useful question is whether relevant customers can find accurate information and take the next step, followed by whether those actions produce qualified business.

Price the work against contribution, not citation counts

Start with your own contribution per completed sale, after the variable costs of serving that customer. Then compare the proposed fee with the incremental completed business needed to cover it. Revenue alone can make an acquisition channel look healthier than it is.

For our own small restaurant example, assume an AI visibility package costs $600 monthly, each additional completed reservation contributes $60, and 5% of attributable visits become completed reservations. Breaking even requires $600 ÷ $60 = 10 additional reservations. At the assumed conversion rate, that requires 10 ÷ 0.05 = 200 incremental visits monthly.

Now assume the restaurant records 40 AI referral visits monthly. The same assumptions imply 40 × 0.05 = 2 completed reservations and 2 × $60 = $120 in contribution. These are illustrative assumptions, not restaurant benchmarks. The proposal needs credible evidence of additional volume, better conversion, or another measurable benefit to close the gap.

For our own larger law firm example, assume a $2,000 monthly project, $4,000 contribution per retained matter, and a 1% visit-to-retained-matter rate. Required expected volume is $2,000 ÷ $4,000 = 0.5 incremental matters monthly, or 0.5 ÷ 0.01 = 50 visits. Fractional matters describe an average over time, not a result the firm can collect each month.

This second model can justify a smaller traffic source because the assumed customer contribution is higher. It fails if inquiries concern services the firm does not provide, prospective clients cannot be reached, or matters would have arrived through another channel anyway. Record those distinctions before attributing all AI-associated revenue to the project.

The September 2026 Similarweb report says AI traffic to US retail sites converted 54% better than non-AI sources by May 2026. That is an industry-specific comparison, not permission to increase either worked example's conversion rate. Measure completed reservations or retained matters directly before changing the economics.

These findings cannot forecast your local lead volume

The supplied evidence combines global traffic panels, US browsing behavior, large prompt datasets and an agency's client observations. None provides a controlled estimate of what a particular Las Vegas business gains by purchasing AI SEO. Ahrefs describes correlation, while the referral studies describe traffic observed across their respective populations.

Sterling Sky's June 2026 report is especially relevant to local businesses, but its agency sample is not a census. It found fewer businesses in AI packs than traditional packs in 88% of tracked markets. That supports concern about reduced inclusion; it does not establish your firm's probability of appearing or the inquiries an appearance will generate.

The same report says ChatGPT referrals equaled 0.1% of Google traffic in the agency's first tracked year and 2% in its second. Those are ratios to Google traffic, not shares of all website visits. Growth in that ratio does not establish that AI referrals replaced lost Google demand.

A vendor may point to the accompanying Semrush June 2026 survey: 81% of organizations integrating SEO and AI visibility reported increased AI traffic or leads, versus 36% managing them separately. That is a reported association from a survey. The supplied extract does not provide a basis for promising either probability to a local owner.

Your measurement can also miss customers who see an answer and later call directly. Treat source attribution as incomplete evidence, and preserve an unknown category. Ask customers how they found you, but do not force an uncertain answer into an AI success story.

Keep a record linking source, landing page, inquiry, qualification and completed business wherever observable. If the uncertainty is which process deserves attention, AI consulting for Las Vegas businesses starts with that operational question. If inquiries arrive but follow-up fails, the measurable problem may belong in business automation rather than search visibility.

Questions owners ask

Is AI SEO worth it for a small business?

It can be, when attributable incremental business covers the cost. The studies establish changing click behavior and growing referrals, not a guaranteed return from buying optimization. Use your contribution per completed customer and observed conversion rate. Keep ordinary search performance in the comparison so the project earns credit for additional results rather than relabeled existing demand.

Does getting cited by ChatGPT mean more website traffic?

A citation creates an opportunity for a visit, but it is not a visit. Semrush shows that brand mentions and cited domains can differ, while Similarweb measures actual outbound referrals separately. Record whether your business was mentioned, whether your domain was linked, and whether a visitor arrived before connecting visibility to customer outcomes.

Should I replace Google SEO with AI SEO?

These studies do not support replacing Google work wholesale. Sterling Sky's observed ChatGPT referrals remained small relative to Google traffic, even after growth. Maintain accurate business information and useful service pages while testing AI visibility alongside them. Change the budget when your own qualified inquiries and completed business justify it, rather than when platform usage headlines rise.

What to do this week

1. Build a baseline from existing records. Export search clicks and recorded AI referrals for the same reporting period. Separate landing pages and inquiry sources, then connect inquiries to qualified opportunities and completed business where possible. Preserve unknown sources instead of treating missing attribution as zero AI influence.
2. Turn the proposal into a break-even calculation. Use the actual fee, your contribution per completed customer and your observed conversion rate. Ask the vendor to identify which assumptions its work can influence. Require mentions, linked citations, visits and customer outcomes to appear as separate measures in reporting.
3. Run a bounded test with a written decision rule. Choose a service with clear customer intent and a working contact path. Save the prompts, dates and answers used to assess visibility, and agree on the reporting period before work begins. Continue only when the evidence supports the cost or identifies a specific correction worth testing.

For help applying this to your business, book a free 15-minute audit through our Las Vegas AI SEO page.