How can I get more traffic from Google and AI search?

Asking how to get more traffic from both channels contains a hidden assumption worth surfacing: that they both produce traffic. One of them mostly does not, and that is the most important thing to understand before planning around it.

Two different kinds of value

Google produces sessions. Someone searches, sees your result, clicks, arrives. Measurable, attributable, and still the large majority of organic acquisition for almost every business.

AI search mostly produces shortlist entries. Someone asks an assistant which tool to use, it names three companies, and the field narrows before anyone visits a website. Direct referral traffic exists — a low single-digit percentage of organic for most sites in 2026 — but it is not where the value concentrates.

The most important thing AI search does for you generates no session at all: it decides whether you are on the list.

Treating them as one traffic number hides that entirely, and leads teams to dismiss the newer channel because its session count is small.

Growing the Google side

Four levers, in order of payback speed.

Click-through rate. Export Search Console, sort by impressions descending and CTR ascending. The top of that list is pages already shown to thousands of people who choose someone else. Rewriting a title is minutes of work against traffic that already exists, and it shows within one to three weeks.

Position. Work the 8–20 band, where the click-through curve is steepest. Moving from 12 to 6 can double a page’s traffic; moving from 60 to 40 changes nothing.

Coverage. New pages, but only against genuine gaps. Publishing into topics you already cover creates cannibalisation rather than coverage.

Demand. Brand work, slow, and now doubly valuable because brand strength is also one of the strongest inputs into whether an assistant names you.

Growing the AI side

Three requirements, in order.

Retrievability. AI crawlers do not execute JavaScript. Vercel and MERJ recorded zero JavaScript execution across more than 500 million GPTBot fetches, with the same behaviour from ClaudeBot and PerplexityBot. If your answer only exists after a script runs, your AI visibility is zero regardless of everything else. Five minutes to check, days to fix, and frequently the entire problem.

Quotability. Answer one question in the first two sentences, in a passage that survives being copied out of context, containing something concrete enough to attribute.

Corroboration. Consistent description across the third-party sources assistants already cite in your category. Slowest, highest ceiling, and the one that requires PR rather than marketing execution.

The work that does both

Five things pay into both channels from a single effort, and they should dominate the roadmap.

• Answering the question early. Wins featured snippets and AI extraction with the same edit.
• Specific, dated, sourced claims. Earns links, snippets and citations. The Princeton and IIT Delhi GEO study found statistics and cited sources among the highest-performing methods for AI visibility, at roughly 30–40 percent relative improvement.
• Consolidating duplicate pages. Improves rankings and produces cleaner citation candidates.
• Original data. The single most durable asset in either channel, and the only one competitors cannot replicate by writing harder.
• Question-led headings. Match how people search and how they prompt.

If a proposed task is not on this list and does not clearly serve one channel, it is probably not the best use of the sprint.

The surface that reduces both

One complication sits between the two channels: AI Overviews take from Google traffic and are not an AI referral channel either.

They appear on roughly 32.5% of monitored queries, in a stable 31–34% band through 2026. Pew Research found users clicked a result in 8% of visits where an AI summary appeared against 15% where one did not. Ahrefs measured a 58% CTR decline for top-ranking pages in February 2026, up from 34.5% in April 2025. And a click from an AI Overview citation arrives as ordinary Google organic traffic, so it is invisible as an AI channel.

Three responses work. Be cited inside the overview rather than listed beneath it, which requires ranking well and owning a clean passage. Shift investment toward query types that trigger overviews least — transactional and navigational. And build content that cannot be summarised away: tools, datasets, communities, and decisions that need your interface.

What growth is realistically available

Before setting a target, establish the ceiling. Total impressions multiplied by a realistic best-case click-through rate gives the honest upper bound of what your current query set can produce on the Google side. If your target exceeds it, the plan needs new coverage or new demand by definition.

Apply a discount for AI Overview presence on the affected subset rather than modelling the whole site at historical rates. On queries where an overview appears, the same position now yields materially fewer clicks than it did two years ago — Ahrefs measured a 58% CTR decline for top-ranking pages in February 2026, up from 34.5% in April 2025.

On the AI side there is no equivalent ceiling calculation, because the outcome is not sessions. The honest framing is share of voice against named competitors, with a target expressed as displacement rather than volume: being named in six of twenty answers where you are currently named in two.

Working out your channel mix

Before allocating effort, establish how much of your acquisition each channel could plausibly produce. Three checks.

Intent mix. Export your top 200 queries and tag each informational, commercial, transactional or navigational. A site that is seventy percent informational is heavily exposed to AI Overviews and has a large AI opportunity. A site that is seventy percent transactional has less of both.

AI Overview presence on your actual queries. Check a representative sample logged out. The industry average of roughly a third is meaningless for your site; your own rate might be five percent or seventy.

Whether your category gets shortlisted. Ask ten buying questions to an assistant. If they return named recommendations, shortlist entry is a real mechanism in your market. If they return generic advice with no companies named, the AI opportunity is smaller and further off.

Those three answers determine whether AI search is a quarter of your opportunity or a rounding error, and they take an afternoon to establish rather than being guessed from industry commentary.

Where the two channels reinforce each other

Three specific loops make the combined programme worth more than the sum of its parts.

Citations produce links. A page cited by an assistant is a page more people read and reference, which produces the backlinks that improve organic ranking, which improves retrieval, which produces more citations.

Rankings produce citations. Google AI Overviews draw candidates from the organic index, so every position gained is a citation opportunity created — for free, from work you were doing anyway.

Assistant mentions produce branded search. When an assistant names you, some proportion of those people search your brand rather than clicking. That branded search is the highest-converting traffic on the site and it is invisible as an AI referral.

None of these loops runs if the channels are managed separately, because the handoffs between them are exactly what a split ownership model drops.

The case for doing both even when one dominates

Even where Google produces ninety-five percent of sessions, the AI work is worth its small allocation, for a reason that is not about traffic.

The retrievability fix costs days and is permanent. The corroboration work is slow enough that starting it two years late cannot be caught up quickly. And the direction of travel is not seriously disputed — Google reported 2.5 billion monthly active users for AI Overviews and 1 billion for AI Mode in May 2026. A small ongoing allocation now buys a position that would cost considerably more to acquire later.

Measuring the combined result

Four numbers, reported together, none of them a blended average.

1. Queries in the top ten. The clearest proxy for Google visibility, hard to game.
2. Non-branded organic clicks. Isolates SEO from general marketing effect.
3. Share of voice in AI answers against two named competitors, across a fixed prompt set.
4. AI referral sessions as their own analytics channel, with conversion rate alongside.

Resist combining these into one score. The two channels move at different speeds and for different reasons, and an index that blends them moves for reasons nobody can attribute.

A combined quarter

Weeks 1–2. Baseline both. Search Console export plus twenty prompts across three engines, run twice. Fix retrievability and crawler access.

Weeks 3–4. Titles and descriptions on the ten highest-impression, lowest-CTR pages. Fastest available return, Google-only.

Weeks 5–8. Restructure the ten pages that should be answering your prompt gaps. Answer-first, one question each, one concrete fact added. Both channels.

Weeks 6–12. Commission and publish one piece of original data. Both channels, most durable.

Weeks 4–12, continuous. Corroboration on the domains your scans show assistants citing. AI-heavy, slowest, so it starts early rather than last.

That sequence produces measurable Google movement inside a month and a readable AI trend by week eight, which is roughly the fastest honest schedule available across both.

Sources

• Pew Research Center — approximately 68,000 real search queries; clicks fell from 15% to 8% of visits when an AI summary was present.
• Ahrefs — CTR impact of AI Overviews; 34.5% decline for position one (April 2025), approximately 58% (February 2026).
• Vercel and MERJ — analysis of 500M+ GPTBot fetches finding zero JavaScript execution; same behaviour for ClaudeBot and PerplexityBot.
• SERP feature tracking, 2026 — AI Overview presence averaging ~32.5% of monitored queries within a 31.02%–34.40% band.

Frequently asked questions