How AI Search Engines Are Changing Website Traffic and Online Discovery

By LinkinGrow. Published October 28, 2026. AI-assisted editorial guide.

For twenty years, search sent people to websites. AI search increasingly answers first and sends people second—if at all. The click is no longer the only proof that a page was useful, and the website is no longer the only place where discovery happens.

This guide explains what has changed in measurable terms, how it affects publishers and online businesses differently, what still holds in SEO, and how answer engines decide which sources to show. It separates documented facts from our analysis, and it does not claim that every site is losing traffic: the effect depends heavily on query type, audience, and what a page offers that a summary cannot.

The short answer

AI search engines—Google AI Overviews and AI Mode, ChatGPT search, Perplexity, Copilot, and others—summarize information directly in the results. For many informational questions, that reduces the need to click. Pew Research Center's 2025 browsing study found Google users clicked a traditional result in about 8% of visits when an AI summary appeared, compared with about 15% when none appeared, and rarely clicked the links inside the summary itself.

At the same time, AI systems create a new layer of discovery: being named, cited, or recommended inside the answer. Fewer, later, better-qualified visits are a plausible outcome for businesses that are well represented; fewer visits with no compensating visibility is the outcome for those that are not.

What actually changed

From ten blue links to a composed answer

Classic search returned a ranked list and left synthesis to the reader. Answer engines do the synthesis: they retrieve several sources, extract passages, and compose a response. Google says its AI features surface links to help people find more on the web, and that the same Search eligibility and quality systems apply. The user, however, sees the summary before any link.

From one engine to many entry points

Discovery now happens across Google, chat assistants, AI browsers, and agents. ChatGPT search fetches live web results and shows sources; OpenAI documents a separate crawler, OAI-SearchBot, for that surface. Each entry point has its own source selection, citation format, and referral behavior.

From crawled-and-clicked to crawled-and-summarized

Cloudflare's analysis of its network reported that AI crawlers fetch far more pages than they send back as referral visits—a gap much wider than for traditional search crawlers. Exact ratios vary by provider and month, but the direction is consistent: content is consumed more often than it is visited.

Where traffic is shifting

The impact is uneven. Based on public studies and our own observation of answer sessions, the pattern looks like this (analysis, not a universal rule):

• Most exposed: short definitions, simple how-tos, quick facts, conversions, and “what is” queries a paragraph can satisfy.
• Moderately exposed: comparisons and “best” lists, where answers summarize options but users still click to verify or buy.
• Least exposed: transactional, local, account, and tool-based tasks; original data; live information; and experiences a summary cannot replace.

Branded search and direct visits often become more important, because a user who first meets a company inside an AI answer may later search for it by name or type its URL. That influence rarely appears as an “AI referral” in analytics.

Publishers

Publishers that monetize page views carry the most direct risk. When an answer reproduces the useful core of an explainer, the page loses the visit and the ad impression. Three responses are emerging:

• Publish what summaries cite but cannot replace: original reporting, datasets, benchmarks, primary documents, and expert analysis.
• Build direct relationships: newsletters, memberships, apps, and communities that do not depend on a referrer.
• Make deliberate access decisions: providers document separate controls for search and training crawlers in some cases. Blocking a search crawler can remove a site from that answer surface; it is a business choice, not a technical fix.

Citations still matter for publishers even when clicks fall: they build recognition among readers, journalists, and researchers, and repeated citation makes a source more likely to be found again.

Online businesses

For ecommerce, SaaS, and service companies, the critical question shifts from “Do we rank?” to “When a buyer asks an AI for options, are we included, and are we described accurately?” A buyer may shortlist three vendors in a chat session before visiting any website.

Practical priorities: keep product, pricing, availability, and policy pages accurate and crawlable; make comparison-relevant facts explicit; maintain consistent business details across your site and reputable profiles; and earn independent coverage that answer engines can cite. Product and local information that systems can verify tends to travel further than marketing copy.

What it means for SEO

SEO is not obsolete; its foundations are the entry ticket. Google states there are no extra requirements to appear in AI features beyond being indexed and eligible for a snippet, and that its helpful-content guidance still applies. What changes is the goal and the scoreboard:

• Rankings remain a leading indicator, but position one may sit beneath an answer that absorbs most attention.
• Clear, self-contained passages that answer a specific question are easier to extract and cite.
• Entity clarity—who you are, what you offer, where, for whom—matters as much as keywords.
• Off-site evidence (reviews on independent platforms, press, documentation, community discussion) increasingly shapes how a brand is described.

This is why disciplines like answer engine optimization (AEO) and generative engine optimization (GEO) have emerged: they extend SEO from earning a ranked link to earning inclusion in a composed answer.

How answer engines choose sources

No provider publishes a complete ranking formula. From documentation and repeated observation, source selection appears to depend on a few broad conditions: the page must be accessible to the relevant crawler; it must be retrieved for the query (often via a search index); it must contain a passage that directly supports a claim; and the source must appear credible relative to alternatives. Pages that hide answers behind scripts, interstitials, or vague copy tend to lose to pages that state facts plainly.

Answers also vary between sessions, phrasings, locations, and model versions. A single screenshot is an anecdote; a repeated sample is evidence.

Measuring the new funnel

1. Referral traffic: segment visits from AI referrers (for example chatgpt.com, perplexity.ai, copilot.microsoft.com) and track their conversion rate, not just volume.
2. Search Console: watch impressions versus clicks on informational queries; rising impressions with falling CTR often signal answer absorption.
3. Answer sampling: run a fixed set of buyer questions across engines on a schedule and log mentions, citations, recommendations, and accuracy.
4. Branded demand: track branded search and direct traffic as a proxy for unattributed AI influence.
5. Self-reported attribution: add “How did you hear about us?” with an AI assistant option to lead forms.

A fictional worked example

Consider a hypothetical Texas outdoor-furniture retailer. Over a year, its “how to clean teak” guide loses most of its clicks as AI summaries answer the question. Yet its sampled visibility for “best weatherproof patio sets for humid climates” improves after it publishes a test-based materials comparison and earns coverage on two independent home sites. AI referrals stay small but convert well, and branded searches rise. Total sessions fall; qualified pipeline does not. The lesson is not that traffic is irrelevant, but that traffic alone stopped describing performance.

A 90-day response plan

• Days 1–30: audit crawler access, identify pages losing clicks to summaries, define 20–50 buyer questions, and take a baseline answer sample.
• Days 31–60: fix factual gaps, add direct answer passages, publish one or two pieces of genuinely original evidence, and align entity details across profiles.
• Days 61–90: resample, compare against baseline, review AI referral conversions, and decide where to invest next.

Common questions

Are AI search engines killing website traffic?

They are reducing clicks for many informational queries, as Pew's data indicates, but the effect varies by query type and site. Transactional, local, and original-evidence pages are generally less exposed.

Should publishers block AI crawlers?

It depends on the business model. Blocking training crawlers and blocking search crawlers have different consequences; blocking a search crawler can remove you from that engine's answers. Review each provider's documentation before deciding.

Is SEO still worth investing in?

Yes. Indexing, crawlability, and quality remain prerequisites for appearing in Google's AI features and influence what other engines retrieve. The work expands rather than disappears.

How do I know if AI answers mention my business?

Sample them: ask the same buyer questions across engines repeatedly and record the results. Analytics alone will undercount AI influence.

Can anyone guarantee AI recommendations?

No. You can improve access, accuracy, and evidence, and measure outcomes. No one controls how each system selects and composes answers.

Sources, scope, and disclosure

This guide was produced with AI assistance using the sources below. Statistics are attributed to their publishers and reflect their methodology and period; the traffic-exposure pattern, recommendations, and plan are LinkinGrow's analysis. The retailer is fictional. No original benchmark or client outcome is claimed.

1. Pew Research Center: Google users are less likely to click on links when an AI summary appears (July 2025)
2. Google Search Central: AI features and your website
3. OpenAI: Overview of OpenAI crawlers
4. OpenAI Help: ChatGPT search
5. Cloudflare: The crawl-to-click gap
6. Google Search Central: Creating helpful, reliable, people-first content