When Search Stops Sending: Designing for an AI-First Google

You open Google, type a question - and just as you finish the keystroke, the answer’s already there.

No clicking. No visiting. No need to meet the site that wrote it.

For the user, it’s convenient.

For the people who build websites, it’s a quiet recalibration of everything.

Because if Google now summarises your work, and doesn’t send people to it, what happens to all the things we make?

If a designer builds a page no one visits, is it really a page?

That’s the question this piece wants to sit with.

That’s online life, after the blue links.

When the web stops being visited

Let’s open with a number. In early 2025, Pew Research found that when an AI summary appears in Google’s results, users click on a traditional result only 8 % of the time.

Without the summary? 15 %. Nearly halved. Ouch…

Analytics from SimilarWeb show that for news-related queries, zero-click traffic has jumped from 56 % in mid-2024 to nearly 69 % by May 2025.

For many queries, the user gets their answer before they ever reach you.

The bottom line: your site’s front door will probably see fewer visitors. So is it worth repainting that front door?

Maybe. But the welcome mat now says Google, in blue, red, yellow and green.

The economic reality

Traffic is oxygen for the web, and the tank is running low.

Ad impressions, affiliate clicks, sponsorships - they all breathe it in.

When the supply drops, the ecosystem changes.

• Ads: fewer sessions → fewer impressions → weaker RPMs (Revenue Per Mille).
• Affiliates: if the AI already answered, there’s no “assist click”.
• Sponsorships: brands now want guaranteed reach - newsletters, podcasts, communities - over search exposure.

One striking data point: visits to news sites reportedly fell from 2.3 billion in mid-2024 to 1.7 billion by mid-2025 (New York Post). (Yeah, I know it’s the New York Post)

The design lesson? Discovery via search is no longer a certainty.
Design for returning visitors and direct relationships.

Offline businesses have always known it: return business is good business.

Designing for machines (without designing for machines)

If your site wants to be quoted, cited, and clicked from an AI summary, you need to make it legible to both people and machines.

1. Structure your content like data.

Use semantic markup (Article, FAQPage, HowTo, Product) and make sure your structured data mirrors visible content. Validate with Google’s Rich Results Test.

2. Answer first, nuance second.

Lead with the distilled answer - 2 to 4 sentences. Then add texture, examples, or the story.

Summarisers tend to lift whatever appears first.

3. Be auditable.

Use clear bylines, “last updated” dates, and citations.

Avoid screenshots of text or charts - use tables and alt text.

4. Scannable beats poetic.

Short paragraphs. Descriptive sub-heads. Bullet points.

Design for tired humans and fast-scanning algorithms.

(To be honest, this makes me sad. Language is a blessing.)

5. Use snippet controls wisely.

nosnippet or max-snippet:0 might “protect” your copy, but can also remove your page from visibility entirely.

Use data-nosnippet selectively.

See WordStream’s guide for examples.

6. Keep your promise.

If your title says Definitive Guide, your first screenful should behave like one.

Both humans and AI models penalise bait-and-switch, as they should.

Examples: turning the ship

Not every business has taken this quietly. Some have found ways to bend the curve, if not fully reverse it.

Real-world: ClickUp’s content refresh comeback.

ClickUp, the productivity-software platform, saw organic traffic slide after Google’s generative changes began surfacing AI-written summaries ahead of search results.

Instead of chasing volume, they re-invested in content refreshes: rewriting intros in a first-person voice, adding comparison tables, and weaving in verified user reviews.

Just four refreshed articles delivered an extra 600,000 monthly clicks within six months.

The lesson: depth, authenticity, and structured clarity beat thin repetition.
 (Read the case study →)

Hypothetical: adapting in action

We aren’t all working for large organisations, here’s a play for something more modest.

A mid-sized publisher noticed that two “How to” articles lost 35 % of search referrals after AI summaries appeared for their target phrases.

They responded by:

1. Adding a crisp, factual answer block at the top.
2. Embedding a small calculator tool that the AI couldn’t reproduce.
3. Inviting readers to a weekly email series.
4. Tracking branded search and newsletter growth instead of raw traffic.

This was, of course, a design shift too; they moved from narrative pages to utility pages.

Three months later: branded search up 20 %, email sign-ups up 25 %, steady engagement.

Clicks stayed lower, but recognition climbed.

The lesson: trade visitor quantity for quality - and come out ahead.

Differentiating in an AI-saturated SERP

AI compresses the average. You win by being the exception.

• Original data: Publish your own benchmarks or studies.
• Interactive tools: Calculators, checkers, visualisers.
• Community: Newsletters, forums, live sessions.
• Voice: Distinct tone and personality still travel.
• Freshness: Visible update dates improve inclusion.

AI eats the average - ship what’s scarce.

Analytics after the click (or not)

When clicks no longer tell the whole story, it’s time to look elsewhere for signals.

Track impressions & appearances

Use Google Search Console → Performance → Impressions to track how often your pages appear.

If impressions rise but clicks don’t, visibility is growing even if traffic isn’t.

Measure assist metrics

• Branded search: track via Search Console’s query filter (contains: your brand name).
• Direct visits: view in Google Analytics → Traffic Acquisition → Session source/medium.
• Newsletter sign-ups: integrate with GA4 events or your email platform metrics.
• Scroll depth & dwell time: use GA4’s scroll and engagement_time_msec events to gauge on-page interest.

Build attribution bridges

Add simple “How did you find this?” polls (e.g., Typeform embed).
Use channel-specific UTM codes like
 ?utm_source=newsletter&utm_medium=email&utm_campaign=ai-summary-article.

Compare conversions by source.

Run experiments

Create twin pages - one answer-first, one legacy.

Track for 4-6 weeks: impressions, CTR, engagement.

Keep whichever performs better.

Clicks are lagging indicators; recognition is leading.

Rights, robots and responsibility

Old controls like robots.txt and meta-robots still matter, but they don’t fully govern AI reuse.

• Snippet controls: tags like max-snippet, nosnippet, data-nosnippet affect reuse - but can also nuke visibility. (Digiday explainer)
• Reuse preferences: Cloudflare Content Signals lets you express reuse intent (“search”, “ai-train”), though adoption is early.
• Choose your posture: public info open; proprietary or premium content guarded.
• Be transparent: visible bylines, update logs, usage terms.

In short, design with intent - both for humans and for whatever’s crawling next.

A 90-Day Survival Plan

If you’re wondering where to start, here’s a three-month stabilisation plan.

Weeks 1-2:
Add clear “answer blocks” to top pages.

Weeks 3-4:
Audit structured data; fix mismatches.

Weeks 5-6:
Create one AI-resistant asset - tool, dataset, or report.

Weeks 7-8:
Strengthen expertise signals (bylines, dates, transcripts, alt text).

Weeks 9-10:
Build a direct channel (newsletter, RSS, community).

Weeks 11-12:
Review impressions, branded search, and engagement.
Optimise for visibility and recall, not just raw traffic.

Things worth remembering

• Hallucinations & trust: AI can be confidently wrong. Transparent, cited pages are more quotable. (The Times)
• Regulatory tension: EU and UK publishers are already challenging attribution gaps. (The Guardian)
• Accessibility as strategy: semantic HTML, alt text, transcripts - great UX for humans, perfect context for AI. Accessibility isn’t just empathy -it’s strategic metadata.

Closing thoughts

The web is evolving from a network of destinations to a network of descriptions. This is actually a pretty monumental change.

But the fundamentals haven’t changed: clarity, trust, structure, and voice still win.

Design for humans.

Format for machines.

Build things worth quoting - even when the click never comes.

In the days of blue links, we fought for clicks. In the days of AI-first discovery, we’ll fight for citations, signals and trust. Will your page still exist if no one ever visits it - or will it live in answers instead?”

If you enjoyed this piece

Follow me for more articles about design-driven coding, usability, and the messy, wonderful middle ground between aesthetics and logic.

Author’s note:

I am currently deeply interested in using AI to generate both visual and text-based content. I am actively collaborating with AI on multiple platforms to explore my thoughts on what creativity is and is not.

My current approach is to collaborate with AI by using the output as a foundation upon which to build and modify.

Sources

Pew Research Center
Seroundtable / SimilarWeb
New York Post
Search Engine Land
Digiday
WordStream
The Times
The Guardian