Search and AI statistics in 2026: what the numbers actually say
Most of the search and AI statistics circulating in Quebec are accurate. What they almost always lack is scope: the territory measured, the sample size, the currency and the date. This page takes seven widely cited numbers and gives, for each one, the primary source and what the number does not say.
Sources last verified: September 10, 2026. This page is reviewed every quarter
- The 68% of zero-click searches is a U.S. measurement. No equivalent data exists for Canada or for Quebec.
- Gartner's prediction was about search volume, not traffic. On the metric it named, it turned out to be wrong.
- The most cited study on AI summaries contains a third number nobody repeats: a page with no summary at all gets more than double the click-through rate of a cited page.
- The average cost per click of $5.42 is American, in U.S. dollars, and a median across 23 industries. No public benchmark exists for Quebec.
- Faced with any search statistic, four questions are enough: who measured it, in which territory, on what sample, at what date.
On this page
Scope of a statistic
The scope of a statistic is the set of conditions under which it was measured: the territory covered, the population or sample observed, the period, the currency when money is involved, and the definition used for what is being counted. A number repeated without its scope stays accurate and becomes unusable, because nothing tells you any longer whether it applies to your situation.
Why this work had not been done for the Quebec market
The numbers circulating in presentations and publications in Quebec almost all come from American studies. They are generally transcribed correctly. What gets lost along the way is the footnote: the sample, the country, the currency, the definition of the term being measured.
That loss is not trivial. An advertising budget built on a U.S. median cost per click in U.S. dollars starts with a gap no optimization will close. A decision to shut down a blog, based on a traffic decline measured in the United States, is made without knowing whether the phenomenon has the same magnitude here.
The seven entries that follow use the same format: the number as you hear it, the dated primary source, the real scope and what the number does not say. None is presented to contradict anyone. They are all accurate at the source.
68% of Google searches end without a click
The number is accurate and recent. SparkToro measured 68.01% zero-click searches in the United States over the first four months of 2026, against 60.45% in 2024. Out of a thousand searches, only 276 now lead to the open web.
- The source: SparkToro, study published June 9, 2026, based on Similarweb clickstream data.
- The scope: the United States only, over the first four months of 2026, excluding Google's mobile app.
- What is missing: the direct comparison with 2024 is fragile, because the data provider changed between the two measurements. Rand Fishkin says so himself in the study. And nothing comparable exists for Canada or for Quebec.
What the number allows a Quebec business to conclude: the direction is solid, the magnitude is not. Your own share of zero-click searches can be read in your Search Console, by comparing your impressions and your clicks on your queries. That reading is about your market.
Gartner's prediction of a 25% drop
On February 19, 2024, Gartner published a prediction announcing a 25% drop by 2026. The press release speaks of “search engine volume”, meaning search volume and not traffic to websites. Everything else hinges on that distinction.
- The source: Gartner press release, February 19, 2024.
- The scope: a prediction, never a measurement, on a metric named in the text: search volume.
- What happened: in the second quarter of 2026, Alphabet reports that AI Mode has passed one billion monthly users, that generative features increase the total volume of queries, and that search revenue is up 17% year over year, at US$63.27 billion.
The prediction is therefore wrong on the metric it named: search volume went up. What actually fell is the click through to publishers, and Gartner was not talking about that. Citing this 25% to justify a content budget means relying on a forecast the facts have contradicted.
A prediction from an analyst firm is not data. It has the same value as a well-argued opinion, and it gets checked two years later like any opinion. This one has been checked.
Being cited in an AI summary brings 35% and 91% more clicks
Both numbers are accurate. They are click-through rates, not clicks, and the study that produced them contains a third number almost nobody repeats.
- The source: Seer Interactive, September 2025 update, on 3,119 search terms, 42 organizations and 25.1 million organic impressions, between June 2024 and September 2025.
- The three numbers for the third quarter of 2025: a page on a query with no AI summary gets a 1.45% organic click-through rate. A page cited in a summary gets 0.70%. A page present on a query with a summary but not cited gets 0.52%.
- What is missing: all three groups are down year over year, by 46.2%, 49.4% and 65.2% respectively. The citation limits the loss, it does not reverse it.
The ratio of 0.70 to 0.52 gives exactly the 35% that circulates. The number is therefore correct, and it compares two degraded situations. The best scenario in the study remains the query on which no summary is triggered.
What that changes for an editorial trade-off is covered in our note on what the study really says about citation.
An average cost per click of $5.42
The number is accurate and it is American. Yet it regularly serves as a benchmark in Quebec budgets, in Canadian dollars, with no conversion and no market adjustment.
- The source: WordStream and LocaliQ, 2026 campaign benchmarks, updated May 19, 2026.
- The scope: more than 13,000 U.S. search campaigns, 23 industries, between April 2025 and March 2026, in U.S. dollars, expressed as medians.
- What is missing: the gap between industries is larger than the average itself, from $1.63 for arts and entertainment to $9.87 for legal services. No public benchmark exists for Quebec.
The only usable reference for your budget is your own history and, failing that, the Keyword Planner estimates for your queries and your area. The method for building a budget from those values is detailed in our analysis of advertising spend.
Traffic coming from AI converts three times better
The number is accurate, and the same study contains the detail that sets its real scope.
- The source: Microsoft Clarity, study published November 6, 2025, on 1,277 domains, mostly publishers and news sites, with one month of conversion data.
- The scope: the conversions measured are mostly sign-ups and subscriptions, not product sales. A little over half of the domains observed recorded at least one during the month.
- What is missing: Microsoft writes in the same study that referrals coming from AI represent less than 1% of total traffic. A channel that converts three times better on less than 1% of the volume remains a small channel.
The practical conclusion fits in one sentence: this channel deserves to be measured separately starting now, using the measurement method we apply on a GEO engagement, and it does not yet deserve a budget reallocation. What online retail is seeing on its side is covered in our analysis of AI traffic in online retail.
57% of the web is bot traffic
The number is accurate, and it applies to a specific network, with a specific definition.
- The source: Cloudflare, announcement of June 3, 2026, on HTTP requests for HTML content passing through its network, at 57.3% against 42.7% for people.
- The scope: Cloudflare's network, not the entire web. The executive who announced the measurement himself calls it approximate.
- What is missing: Fastly's 37%, often cited in the same sentence, comes from another network, another definition and another year, in August 2025. The two numbers cannot be compared, and the gap between them measures no progression.
What this changes for you can be read in your logs, not in the global number. The share of bots on your own site, the agents that make it up and what they cost in hosting are covered in our note on AI crawlers and the hosting bill.
Orders coming from AI have tripled at Shopify
The number is accurate and it is relative. Its relative nature is what explains the differences between the versions in circulation.
- The source: Shopify's results for the second quarter of 2026, published August 5, 2026: traffic referred by AI tripled year over year, and orders coming from those channels tripled as well.
- The scope: year-over-year growth across all the stores on the platform, all sizes and all markets combined.
- What is missing: Shopify publishes no absolute value, so the starting base remains unknown. The higher multiples in circulation come from a November 2025 communication, on another period and another base.
A tripling on an unknown base can represent a significant channel just as it can represent a fraction of a percent. The same quarterly report indicates that classic search accounts for about a third of store traffic, which gives the scale.
The four questions to ask about any search statistic
These four questions are enough to know whether a number applies to your business. They take one minute to ask, and they rule out the majority of misused numbers.
- Who measured it: a platform on its own data, a tool on its panel, or a firm making a prediction.
- In which territory: the United States, Canada, or a global aggregate dominated by a single country.
- On what sample: how many sites, queries or campaigns, and of what type.
- At what date: the period measured, not the publication date of the post that repeats it.
A vendor who answers all four questions without hesitating has read the source. One who points you to the blog post that cites the study has not read it.
Knowing which of these numbers apply to your market and which do not is the starting point of a 90-minute consultation.
On Quebec engagements, we redo these measurements on the client's data rather than reusing the American benchmarks. The share of zero-click searches is calculated in Search Console from impressions and clicks on the queries that matter. The reference cost per click is taken from the advertising account's history, over the last twelve months, in Canadian dollars. The share of traffic coming from assistants is read in the corresponding GA4 channel, knowing that this channel excludes Google's generative surfaces and therefore underestimates the phenomenon. What stays in-house is the inquiries actually received over the period: it is the only curve that tells you whether a traffic decline cost you anything.
What these seven numbers say together
Taken one by one, these numbers describe different phenomena. Taken together, they draw a single trend: search volume is rising, clicks through to websites are falling, and the new channels convert well on very small volumes.
That reading has one budget consequence, and it is the only one that can honestly be drawn from the whole. Traffic stops being a measure of success you can use on its own, because it falls for reasons that say nothing about your sales. The indicators that stay readable are inquiries received, cost per inquiry and the share of your commercial queries where you appear. How to build that dashboard is detailed in the SEO indicators worth tracking.
The measured effect of AI summaries on traffic is covered in our analysis of generated summaries and the traffic decline. Building a marketing budget that accounts for these shifts is covered in our analysis of the marketing plan and its budget.
Rebuilding a measurement that holds up on your own data is at the heart of the Attract customers with SEO and AI goal.
Already running a marketing team? See how we plug in as reinforcement on organic search.
Frequently asked questions about search statistics
What percentage of Google searches end without a click?
68.01% in the United States over the first four months of 2026, according to SparkToro, based on Similarweb clickstream data. The number excludes Google's mobile app. No equivalent measurement exists for Canada or for Quebec.
Is it true that search engine traffic is dropping by 25%?
No, not on the metric announced. Gartner's prediction of February 19, 2024 was about search volume, and that volume has increased: in the second quarter of 2026, Alphabet reports more than one billion monthly AI Mode users and search revenue up 17%. What has fallen is the click through to websites, which is a different measure.
Is it worth being cited in an AI summary?
Yes when a summary is triggered, because a cited page gets a 0.70% organic click-through rate against 0.52% for a page that is not cited. The best scenario, however, remains the query with no summary, at 1.45%. These three values come from the Seer Interactive study updated in September 2025.
What cost per click should you plan for Google Ads in Quebec?
No public benchmark exists for Quebec. The $5.42 often cited is a U.S. median in U.S. dollars, across 23 industries, published by WordStream and LocaliQ on May 19, 2026. Your account history and the Keyword Planner estimates for your area are the only two usable bases.
Does traffic coming from assistants really convert better?
It converts about three times better according to the Microsoft Clarity study of November 6, 2025, conducted on 1,277 domains, mostly news sites. The same study specifies that this traffic represents less than 1% of total volume, which makes it a channel to measure rather than a channel to fund for now.
How can you verify a search statistic yourself?
By going back to the primary source rather than the post that cites it, then noting four elements: who measured it, in which territory, on what sample and at what date. A number for which those four elements are not available should not be used to decide a budget.
- SparkToro, zero-click search study, June 9, 2026, Similarweb clickstream data, United States.
- Gartner, press release of February 19, 2024 on the evolution of search volume.
- Alphabet, second quarter 2026 results, published July 22, 2026.
- Seer Interactive, AIO impact on Google CTR, September 2025 update, 3,119 terms, 42 organizations.
- WordStream and LocaliQ, 2026 search campaign benchmarks, updated May 19, 2026, U.S. campaigns from April 2025 to March 2026.
- Microsoft Clarity, AI traffic converts at 3x the rate of other channels, November 6, 2025, 1,277 domains.
- Cloudflare, announcement of June 3, 2026 on the share of automated traffic, HTTP requests for HTML content on its network.
- Fastly, research on automated traffic, August 19, 2025.
- Shopify, second quarter 2026 results, published August 5, 2026.
Geneviève puts the strategy for your engagement into action. She leads all our web development projects: Shopify, WordPress and the new ways of building a site with AI. She manages our team of developers and translates your business needs into technical language. She runs your organic search (SEO), your visibility in AI answers (GEO) and your site's conversion rate optimization (CRO). Her work is at the heart of three goals: Attract customers with SEO and AI, Improve your site's conversion, and Strengthen your visibility in AI answers. With Gabriel, she also builds the landing pages for your advertising campaigns. She writes mainly about SEO, AI visibility and web design.
About Falia →
SEO
Paid advertising
Conversion
AI visibility
Web design