Browser Wars Are Coming To AI Search: An AMA With Mark-Williams Cook

There were truthful galore hunt changes successful 2025 that it was difficult to support up pinch the industry.

Mark Williams-Cook has been intimately search these changes. In this AMA, he answers burning questions astir declining clicks, caller Generative Engine Optimization tactics, and hunt skills that will beryllium captious for occurrence successful 2026.

1.Google claims AI Overviews thrust “higher quality” clicks. Is that conscionable PR spin, aliases are you seeing grounds of thing other successful the data?

Google says searches are up, astir 20% much queries from 2024 to 2025, but full integrated clicks are “relatively stable.”  That intends less clicks per search, particularly on informational queries wherever AI Overviews reply the mobility directly. 

Google’s declare is that mean click value has improved, not that each clicks are better. In reality, we’re getting less clicks overall, but the ones that select done whitethorn beryllium much valuable. It’s a clever measurement of saying: “You’re not getting much traffic, but what you do get is higher quality.” So yes, there’s decidedly immoderate PR rotation successful really it’s presented.

2. How overmuch does Chrome information support Google’s monopoly and profitability?

Well, the antitrust trial didn’t find it impactful capable to unit them to driblet it. I deliberation Chrome information astir apt plays a bigger domiciled successful PPC than SEO, chiefly successful profiling personification behavior.

What caught my attraction was really the DOJ proceedings lined up pinch the content storage leak. There were mentions of Chrome passing URL fame scores. However, Google insists they don’t usage Chrome information successful rankings, and I thin to judge they’re not lying, but they building things very carefully.

Still, it wouldn’t astonishment maine if Chrome information is utilized to find which URLs should beryllium crawled, indexed, aliases moreover re-ranked based connected popularity. However, I don’t deliberation it’s arsenic powerful arsenic SERP usage data, specified arsenic who clicks what, really agelong they stay, and whether they bounce aliases betterment queries.

Looking ahead, browsers will matter a batch more. LLM hunt platforms often trust connected web hunt successful the background, which heavy depends connected SERP usage data. If hunt moves into LLMs, we suffer that data.

That’s why I deliberation we’re seeing companies for illustration Perplexity and ChatGPT building browsers arsenic a measurement to way what users find adjuvant and to bring agentic AI experiences straight into the browser. 

So yes, browser wars are decidedly coming!

3. What insights do you person astir LLMs ranking for multilingual queries? 

This is simply a basal rumor pinch a exemplary based connected connection alternatively than connected documents delivered successful containers aliases websites that tin supply important context. 

We’ve seen akin issues for years pinch features like People Also Ask because those suggestions are besides language-model-driven. For example, if you do a hunt successful the UK, you’d get answers for illustration Walmart and U.S.-centric results that are wholly irrelevant to section users.

If the AI strategy uses retrieval-augmented procreation (RAG), modular optimizations still apply. But for the guidelines model’s behavior, unless the personification explicitly sets their region aliases context, there's not overmuch you tin do to power it.

Honestly, this feels for illustration a level problem, not a patient problem. Until the AI hunt providers build successful amended contextual awareness, there’s only truthful overmuch we tin do.

That’s the billion-dollar question. I don’t deliberation SEOs person ever afloat nailed what value contented means. We’ve called it unique, helpful, aliases good, but those labels are vague.

At an SEO FOMO event, Gary Illyes said Google will beryllium focusing much connected original content. That’s not new, but what counts arsenic original has changed. 

Ten years ago, it meant rewriting existing info. Then we discussed accusation gain, which involves adding thing new. More recently, it’s astir offering a individual constituent of view, because AI can’t replicate lived experience.

I deliberation of contented successful 2 buckets:

• Solved information:  Slow-changing facts that AI tin confidently generate. If your website relies connected top-of-the-funnel information, expect it to get eaten up by AI answers. LLMs don’t request to execute immoderate grounding aliases retrieval for these queries because they're truthful assured successful their expertise to make the answer.
• Human content: Copy that reflects emotion, opinion, aliases storytelling. I've developed a sixth consciousness erstwhile reference thing that's AI-related, and I instantly consciousness disappointed, disengaged, and don't want to proceed reference it. People are craving thing much human, which is why creators for illustration Ed Zitron person thriving, paid newsletters.

Ultimately, it comes down to your business model. If you trust connected cranking retired high-volume, low-differentiation content, this alteration whitethorn beryllium existential. But for everyone else, it’s an opportunity to do amended activity and extremity rewarding fluff.

5. Is video now a required SEO maneuver to summation visibility successful AI and PAA results?

Video isn’t required, but it’s a awesome starting point, particularly erstwhile moving pinch taxable matter experts who struggle pinch writing. 

We often grounds interviews pinch clients and past use AI to create content based connected their unsocial perspectives. It’s often faster and much aligned than penning from a brief, because you’re capturing what they think, not conscionable answering prompts.

As for People Also Ask, I don’t attraction overmuch connected trying to rank there. I usage it much for understanding search intent. It tells maine what Google expects to spot adjacent a topic, and that’s adjuvant for contented creation.

6. Are we overestimating the threat of reply engines, aliases is complacency dangerous?

Complacency is ever dangerous. Regardless of the industry, the aforesaid money flows from customers to companies online. So, erstwhile we opportunity “threat,” we must specify what it means.

In position of LLM search versus accepted search, strategies specified arsenic integer PR and organization building are becoming progressively important, but they’ve ever been halfway to SEO.

I deliberation the point we mightiness beryllium complacent astir is really we measurement success. You tin win successful AI hunt without generating immoderate traffic. 

For example, I precocious chose a conveyancer to move location pinch by utilizing LLMs to summarize reviews of circumstantial companies adjacent maine and get an overview of what they were bully at, what they weren't truthful bully at, and the prices group had discussed.

 Then, I Googled the company, went to their website, and converted there. I didn’t thrust immoderate LLM postulation for them, but they still won my business.

So, if you're only search postulation from devices for illustration ChatGPT, you're missing the bigger picture. AI is much of an influencer-type lever than a postulation channel, and that’s a captious distinction.

7. What would you stress erstwhile communicating SEO worth to C-suites pinch each these changes?

At Candour, we’ve started sending short play explainer videos to clients, astir 2 to 4 minutes each, and designed specifically for stakeholders. The videos screen cardinal topics for illustration level shifts, usage statistics, and risks for illustration hallucinations that could effect their brand.

The extremity is to drip-feed useful accusation complete time. You’re not going to alteration someone’s mind successful a azygous meeting. You request to break it down, tailor the connection for each audience, and beryllium consistent.  It’s astir planting seeds truthful they travel to the correct conclusions themselves, astatine their ain pace. 

8. What’s 1 2025 hunt inclination you deliberation was full BS, and what should we beryllium paying attraction to instead?

For me, it’s the rebranding of basal SEO nether fancy GEO terminology. Many of the strategies being touted arsenic caller are things we’ve ever done, conscionable renamed.

Take “chunking” arsenic an example. It’s framed arsenic a method breakthrough wherever you building contented truthful LLMs tin digest it much effectively. 

But really, it’s conscionable penning clearly, breaking up topics, utilizing headers, and not dumping monolithic paragraphs into a wall of text. That’s really we’ve been penning for group for years.

Instead of obsessing complete buzzwords, attraction connected the fundamentals. Make contented better, clearer, and much useful. That’s what moves the needle.

9. Is building communities (forums, Discord, Slack) a smarter hedge than competing head-to-head?

Yeah, for sure. When Reddit started dominating visibility, first successful Google and past successful LLMs, I saw galore group saying, “We request to beryllium posting there.” That ever makes maine cringe a bit, because astir of those efforts thin to spell poorly. I’ve been a Redditor for 19 years, and I’ve seen really that usually plays out.

With clients, we ever commencement by identifying wherever their assemblage already spends time. There’s nary constituent launching a Discord if your customers aren’t there. But erstwhile you get it right, community marketing becomes an motor for everything else, aligning pinch wherever hunt is headed.

Historically, hunt has been drifting distant from webmaster-controlled spaces for years. PageRank was 1 shift, and AI is another. Now, it's much astir what the broader net is saying astir you, specified arsenic your sentiment, visibility, and reputation, than astir your site.

Communities thief style that narrative. If your merchandise is good, a organization often forms anyway. However, investing in, growing, and managing it gives you influence. It’s a smart hedge, because moreover if hunt vanished tomorrow, a beardown organization would still beryllium incredibly valuable.

10. Are location immoderate hunt skills aliases devices SEOs should commencement learning now to get up successful 2026?

Here are a fewer things to attraction on:

Learn really ample connection models work

SEOs often attraction connected tactics, usually owed to unit from managers aliases stakeholders. That’s understandable, but knowing really a connection exemplary works, really pretraining happens, really matter is generated, and what influences it offers a competitory advantage.

When you understand the technology, you tin harvester your LLM and trading knowledge to create your ain strategies, alternatively than relying connected recycled advice.

It besides provides protection; knowing the strategy helps you debar strategies that whitethorn look effective successful the short word but are risky successful the agelong term. Most clients want sustainable success, not short spikes followed by penalties.

Follow thought leaders

Follow thought leaders like Dawn Anderson and Britney Muller. Both were cardinal to the SEO world and past transitioned into fields for illustration accusation retrieval and instrumentality learning. That method position is incredibly valuable and will go much important successful the future.