Google I/O 2026’s announcement changed SEO – Here’s What You Need to Know

In this episode, Bill Hunt and Bill Scully break down the structural implications of Google I/O 2026. As search transitions from a referral index into a task-oriented agent engine, websites must evolve from human-centric destinations into highly optimized, fast, machine-readable APIs.

This episode details the rollout of Gemini 3.5 Flash, the implicit eligibility gates affecting search retrieval, and a 5-pillar operational playbook designed to realign your product, engineering, and marketing teams for the agentic web.

Key Takeaways

  • The Website as an API: Web properties are no longer just human destinations; they function as nodes that AI agents query for data, verify for trust, and use to execute tasks without user browsing.
  • Visibility Engineering: SEO must pivot from a creative content accessory into a rigorous technical discipline focused on schema depth, clean architecture, and total machine legibility.
  • Stricter Performance Gates: Because Gemini 3.5 Flash prioritizes hyper-fast, autonomous agent loops, Google is quietly using site speed, minimal JS bloat, and low API latency as strict eligibility gates for AI retrieval.
  • Off-Site Agentic Commerce: The rollout of the Universal Commerce Protocol (UCP) allows AI agents to handle checkouts natively inside surfaces like Gmail or YouTube, making highly accurate product feeds and structured data paramount.
  • The Traffic Reckoning: With AI Overviews triggering up to a 58% drop in click-through rates for top positions, surviving brands must secure secondary footprints across the external data hubs agents actively scan (e.g., LinkedIn, Reddit, YouTube).

Catch the Full Episode

Youtube: https://youtu.be/1j2qpq53SM4

Spotify: https://open.spotify.com/episode/4NMw3AAQrkwuytXsZaK0W4?si=oHmZHXtRRmuS1IsGrGB8VQ

Apple Podcast: https://podcasts.apple.com/us/podcast/google-i-o-2026s-announcement-changed-seo-heres-what/id1755789311?i=1000771907174

Show Note Links

Transcript

Bill Scully (00:00.000)

Welcome to Episode 43 of the International Web Effectiveness Podcast. I’m Bill Scully, and as always, I’m joined by Bill Hunt.

We’re here to help you improve your web presence and manage your global websites more effectively. Today, we’re talking about Bill’s thoughts on Google I/O 2026, what others in the industry are saying about it, and some of the things we’re seeing with our clients.

Before we get into it, I’d like to ask you to subscribe so you never miss an episode. If you’re enjoying the show, tell someone and leave us a review.

Reviews help keep us going and help new listeners find us.

In our last episode, Episode 42, Frustrated by Friction, we talked about why websites kill business when they make users work too hard. Today’s topic builds on that.

Many of the I/O announcements push more friction onto site owners rather than users.

So Bill, how’s it going? It feels like it’s already been a month.

Bill Scully (01:36.082)

It’s been crazy.

I’m sure you’re doing all kinds of vibe coding, learning new things, and experimenting. You’ve got to love what we do.

So today we’re talking about I/O 2026. Was there anything you saw that surprised you? Was there anything you already saw coming? What are your thoughts?

Bill Hunt (02:47.926)

I thought it was really interesting because Google has always alluded to the idea that any digital asset is fair game, whether it’s YouTube or something else.

We’ve already seen this with changes to the SERP structure. Depending on the query, Google may serve ads, local results, or other content. The search landscape has changed.

Now I think they’ve doubled down on that approach. This is the new normal, and we need to start thinking about it that way.

Bill Hunt (03:33.763)

Now Gemini sits in front of all of that.

In a prior episode, we talked about December, when Google first introduced Gemini into the equation. We discussed how the relationship between the user’s query and the site being presented changed dramatically.

We saw this in real data. A site that may have been a great answer before could suddenly become less relevant when evaluated within the broader context Gemini was using.

The example I remember most vividly involved a query about issues with contact lenses, specifically colored contact lenses.

One site had been dominating those rankings. They had been number one or number two for years.

Bill Scully (05:11.130)

Before you get into that, I wanted to mention something.

Elizabeth Reid, Google’s Vice President of Search, said that AI Mode has surpassed one billion monthly users, with queries more than doubling every quarter since launch.

That’s crazy.

Bill Hunt (05:30.863)

Exactly.

And when you factor Gemini into the equation, it becomes even more significant.

I think BrightEdge did a study across nine vertical sectors and found that AI Overviews appeared in approximately 48% or more of the queries they tested.

That’s a pretty interesting development.

Again, this ties into something I was reading this morning while advising an agency. They were presenting a readout, and it reinforced a point I’ve been making for a while.

Bill Hunt (07:36.099)

There’s a whole process happening before you can even be considered as a potential answer.

Honestly, that’s probably an entire podcast episode by itself.

I’ve been talking about these eligibility gates for a long time, and nobody seemed to care. Now I’m finally starting to hear more people discuss them.

People are beginning to recognize that this gating process is happening.

I wrote about this a year ago, and we’ve talked about it before. Remember the example where Absolut Vodka was showing up in response to a question about vodka?

That kind of behavior illustrates what’s happening behind the scenes.

Bill Scully (09:24.510)

Right.

Bill Hunt (09:40.404)

They’re able to surface that answer because of those mechanisms.

The third and fourth points I wanted to mention relate to Google’s product messaging, specifically what they’re doing with DevTools and some of the other initiatives they highlighted.

The main takeaway I noted was that fast and legible are no longer competitive advantages.

They’re simply the baseline requirements for participating now.

Bill Scully (10:52.478)

Right. Yeah.

Bill Scully (11:01.844)

Yeah, exactly. I think everybody is focused on these content blocks and answering questions.

As technical SEOs, we need to be thinking about how you even get considered in the first place.

Speed, JavaScript performance, and other technical factors can determine whether you’re even eligible to be included.

Bill Hunt (11:30.992)

Exactly.

If they can’t get to you, then none of the rest matters.

My fourth point was this: web effectiveness is now critical. We’ve essentially reached the mainstream, but nobody seems to fully recognize it yet.

If we follow your line of thinking, all this structuring content into chunks, the debates about LLMs.txt files, the different MD formats people are pushing, entity graphs, and all of that—

we still have to stop and ask: can they even reach my page?

Bill Hunt (13:42.682)

To mobile first… well, I think there’s another leap after that.

So what would the third leap be? Maybe teams.

We’re now moving from what I’d call AI web 2.0 into an agentic model, which is version three, where bots are interacting with other bots.

Everything we do to prepare for that future needs to be publishable and accessible on the web.

I was talking to someone recently about being “agentic ready,” which is what Google is effectively pushing toward.

And I asked them a simple question.

Can I come to your website today and go through an entire decision-making process?

Take a hotel, for example. Let’s say I just came back from Porto, Portugal.

If I wanted a hotel in downtown Porto, in the old town, with specific features, at a certain price point, and I didn’t want to manually filter everything—

the system should just understand that intent.

Somehow, it should just know what I want.

Bill Hunt (15:43.770)

So much of what we’ve built is still page-level centric.

I’ll give you an example from this week.

I was reviewing a schema for a client. Their agency had implemented @ID and graph-based structured data, moving toward an entity relationship model instead of pure schema.

What was interesting is that when we tested the page, it wasn’t validating properly in Google’s rich results testing tool.

It flagged missing elements like address and other required fields.

But those fields weren’t missing because of bad data—they were structured differently.

They used @ID to point to a home page that contained that information.

So, in their model, the business details existed—just not on the page itself.

The issue is that the tool still treats it like a single-page audit.

It wasn’t smart enough to follow the link and verify the connected data.

Bill Scully (17:17.935)

So you’re saying the page was correct, but the tool was wrong—or that we need to catch up to how AI crawlers interpret this?

Bill Hunt (17:43.533)

In a sense, both.

AI systems are increasingly entity-based, and Google is explicitly framing things around entities.

But a lot of tools still operate at the page level.

So we end up with this mismatch between how systems understand the web and how we currently evaluate it.

Bill Scully (17:59.937)

Right.

Bill Hunt (17:21.652)

The question becomes: should we be placing information like an address on every single page?

Bill Scully (18:13.350)

It probably depends on the page type.

If it’s a hotel location page, then yes—the address should definitely be there.

Bill Scully (18:21.928)

Right.

Bill Hunt (18:39.117)

Because you haven’t nested it as a graph or properly bound the entity relationships.

You’ve got too much code on the page. The address is included just to satisfy the Google Rich Results tool, but the tool is saying you don’t need it because you can use an @ID to connect it elsewhere.

And this is the key point—AI is looking at things in totality.

It understands the graph.

This is how I’ve been trying to explain it to people.

We sell a product. That product has a name, a brand, and is sold by a specific business type. It may have a manufacturer, and that manufacturer has a legal trade name.

All of that together forms the knowledge graph.

But when we think purely at the page level, we lose that structure.

Bill Hunt (20:35.282)

The third part of this is something that came out this week from Google, related to its product and marketplace feed—the conversational elements.

If you look at these conversational elements, they’re designed to answer more nuanced questions like: does this product have an earphone jack? What are similar products?

These are attributes that are often difficult to infer directly from a page, but can be explicitly provided in structured data.

Google is essentially giving us about nine new categories of tags we can append to merchant feeds.

Bill Scully (20:49.739)

Conversational elements? Is that what you’re referring to?

Bill Hunt (20:53.181)

Yes—conversational attributes.

They include things like question-and-answer data, document links, related item groups, and product variants.

For example, you can now add FAQs about a product directly into the feed.

I’ll put the full list in the show notes so listeners can reference it.

Bill Scully (21:08.576)

I’m looking at it now.

So: question and answer, document links, related product groups, item group titles, variant options…

Even popularity rank—an indicator of how popular or highly rated a product is within your catalog.

Bill Hunt (21:30.233)

Within your catalog.

And here’s the thing—nobody else is going to know this level of detail unless the brand provides it.

This is also going to disintermediate some affiliate ecosystems, because they’re often the ones compiling this kind of information when brands don’t provide it themselves.

Bill Hunt (21:50.256)

When Google introduced AI Overviews and suggested users could research products directly, many affiliates pushed back.

Then Google added direct purchasing through its program, and affiliates objected again, wanting to bundle and promote alternatives.

Google’s response was essentially: we can support bundling—just provide the bundle components in the feed.

Bill Hunt (23:10.913)

Most people don’t think this deeply about products like coffee grinders.

They just want a simple outcome.

Personally, I own four grinders, and they all have the same problem—static buildup that makes a mess when I dump the grounds.

That’s the one thing I wanted to know: does it reduce static?

The brand site didn’t mention it, but reviewers consistently said, “Yes, this finally reduces static.”

So why can’t the brand just say that?

Now they do, but only after external feedback surfaced it.

This is the kind of information users actually care about.

Bill Hunt (25:39.214)

This ties into the messy middle.

Almost nobody truly owns the entire customer journey.

And in order to own it, we have to eliminate friction.

From the last episode—we talked about friction.

You want people to prepay for something like sprinkler maintenance, but you’ve made the process so complex that it becomes difficult to complete.

So how does a bot handle that if a human struggles with it?

If we believe the future narrative, users will delegate these tasks to assistants or agents.

But those agents still need to complete structured steps.

If a human can’t navigate it easily, how will a bot?

That’s the real question: can they get through the journey?

Bill Scully (26:53.780)

Right.

Bill Scully (27:30.388)

Yeah, I think it goes back to our core thinking. Think like an engineer.

How would you handle the ingestion of information and the way it’s displayed to users so that it actually helps them?

Then check what’s happening on your site and figure out why.

But earlier you mentioned something really interesting—someone had lost their schema.

Bill Hunt (29:34.424)

Right.

Just click the little button and look at the source. There’s schema there—so why can’t Google read it?

Then run a Lighthouse or Core Web Vitals test, and suddenly you see the issue.

You’ve got a low performance score because the site takes forever to load.

There are heavy animations, large graphics—like a massive header image—and multiple tracking scripts all firing at once.

And that’s where the real problem is.

Even in cases where schema is injected server-side, it’s present in the raw HTML.

But the HTML is so cluttered that it becomes difficult to process.

It’s technically there, but everything around it prevents it from being effectively read.

And because systems aren’t willing to wait, they just move on.

Bill Hunt (31:00.141)

That’s when teams realize they need to fix it.

Can we lazy-load some of this? Can we reduce what’s blocking rendering?

And that changes the entire dynamic.

Yes, there’s pain and disruption from GEO thinking and AI-driven ranking changes in our industry.

But the upside is that it gives us leverage.

We can finally point to it and say, “This is affecting us from a GEO perspective,” and that moves things internally.

Bill Hunt (31:29.537)

I think we’re finally going to be able to rein in content teams and shift websites toward functioning properly for users—

rather than optimizing purely for awards or recognition like a Webby or a Clio.

Where the goal is not “is this a beautiful piece of content,” but “does this actually work in the system?”

That’s where I think we’re heading.

Bill Scully (31:55.679)

Yeah, we need to stop funding SEO as just content production.

Instead, we should fund it as engineering that controls how you show up in AI systems.

Bill Hunt (33:01.931)

The CEO of Netflix was once asked whether they would shift from long-form content like dramas and films into short-form reels.

And his response—just short of an explicit rejection—was essentially, “Absolutely not. That’s not what we do.”

They’ve designed their system around long-form storytelling.

Bill Hunt (33:43.720)

What we’re seeing now is that these systems are effectively running in sequences—almost like lazy-loaded logic chains.

They progressively build context, layering more information to anticipate future questions.

That’s why understanding speed and performance has become such a critical gate.

Bill Hunt (35:20.053)

In one case, an AI Overview returned results in German about choosing between two shoes from the same brand.

There were no direct product links.

Citations were on the side, mostly from sources like Runner’s World and similar publications.

Then I ran the same query in ChatGPT.

It produced a similar recommendation-style response and included a mix of English and German product references.

But interestingly, none of the products it recommended were actually in the original AI Overview’s ad block.

Bill Hunt (36:35.607)

I intentionally didn’t include links because I wanted to see where it would surface alternatives.

And it ended up showing me discounted listings instead of the official brand site.

It even inferred that I was price-sensitive and prioritized cheaper options over direct brand pages.

So instead of sending me to the manufacturer for research or purchase, it routed me to third-party sellers offering discounts.

Bill Scully (37:09.952)
That’s interesting. Do you think that was accurate, not hallucination?

Bill Scully (37:19.848)
No, no. That’s the reason why he did it?

Bill Hunt (37:24.237)
I think so, right? Because again, it gave me below the two shoes, you know, it said these two are here, are perfect for you.

Below it, it gave me other options, really detailed options.

Bill Scully (37:25.333)
Yeah.

Bill Scully (37:36.864)
Hmm.

Bill Hunt (37:37.316)
But there were no links to anybody to buy it. I thought that was fascinating. Why wouldn’t you throw me a bone?

And I had to ask it, but that’s when it said I didn’t put the links to those product pages because I thought you wanted to buy discounters, and it had discounters listed, right?

So I was like, was it on purpose? Was it an accident? Was it some magical thing that the brand site wasn’t optimal? I don’t know.

Bill Scully (38:02.884)
It makes sense when you think about it from a searcher intent point of view.

Bill Hunt (38:46.359)
It had an article about its shoes and how to pick its shoes and why, for amateur or first-time marathon runners, these would be good, using all kinds of words.

Then I found another version which had this really lofty tone. It’s like, for the urban runner who’s getting into the lifestyle of running.

Bill Scully (39:09.639)
Yeah.

Bill Scully (39:09.639)
Yeah.

Bill Hunt (39:50.066)
Vector relevance score. And the one with the lofty marketing speak—about Burnett being in running in the urban oasis, a cool breeze in your hair, and the euphoric nature and all this happy, happy stuff—was like a 0.25.

So when we’re talking about this retrieval thing, the very first step, which one is it going to bring back?

Bill Scully (40:00.705)
Just do it.

Bill Scully (40:30.111)
Yeah.

Bill Hunt (40:43.449)
But the English version for Germany of the exact same page, all of those phrases had been removed and replaced with “running shoe.”

Because my guess is whoever the localizer was, their brief said, “Tune this page for running shoes.”

So all of those words that made the local market version brilliant—

Bill Hunt (41:36.591)
Relative content. That’s true, because if these are all clones, it can negate them and take the strongest one. So it may take your US page over, and we’ve seen this time and again, even over your German page. It will take the US page and show it because it is the best page.

Bill Scully (41:55.593)
And translated into German for German users, yeah.

Bill Hunt (42:00.524)
Exactly. And all of this is happening because of tools like Gemini. They’re all really getting in there. And I think, you know, that’s what it is. And as more AI mode and generative becomes the new search box—I mean, like you said, it’s over a million queries and doubling every quarter—all of that is really…

Bill Scully (43:10.868)
Well, first why don’t you tell everybody what the conference is and what your title is?

Bill Hunt (43:14.605)
Yeah, it’s called “The Challenges and Mistakes of Global Search: The AI Edition.”

So they wanted me to do my normal rant where I’m complaining about all the stuff we do wrong. Normally it’s like 30 slides on hreflang mistakes and geo-targeting mistakes.

Now it’s like 14 slides of everything that’s a problem before you can even be in the consideration set to even be in the results.

Bill Hunt (43:58.311)
Yeah, the query fan-out.

Bill Scully (44:02.464)
Go ahead, start over.

Bill Hunt (44:29.839)
Twenty-eight, twenty-six sub-queries in the fan-out using Mike King’s Q4. Twenty-eight of them.

And the second one, there were four—I’ve never seen this many reformulations—but my simple query was “best running shoe for first marathon in German.”

There were four sort of refactoring or restatement of that, and one was English.

Bill Hunt (45:25.825)
Of a first marathon and the best running shoes for that.

So I think this is a whole other thing. Gemini is using this too, and this is actually Gemini, because it’s AI overview.

AI overview gave me nine query fan-out sub-prompts. AI mode gave me twenty-six.

Bill Scully (45:50.910)
Wow.

Bill Hunt (45:55.208)
Twenty-six refinements of a simple question: “What was the best running shoe for the first marathon.”

And then I found a plug-in where you can do the same thing in ChatGPT, and it was a similar breakdown. It doesn’t give you as clean a query fan-out, but it gave me all the pages it cited in the categories.

And there was, other than brand-to-brand or product-to-product comparison…

Bill Scully (46:43.968)
It’s entity expansion.

Bill Hunt (46:53.835)

In digging, I’ve been doing a lot of research into tokenization and semantic compression across languages. What I found is that if there’s a word like “first marathon,” in German that becomes a compound word. “First” and “marathon” get combined, and that word doesn’t occur very frequently.

So one of the—then I started digging into this, and there’s actually a really interesting research paper. The math is pretty intense, but if you can get through it, it explains that you need at least twenty to fifty references. Not just you, but the market needs twenty to fifty references of that term for it to really get its own vector.

Otherwise, it gets split vectors or becomes an unknown vector.

Bill Hunt (48:08.497)

So Meta was essentially asking how they can clean up their corpus, tokens, and vectors. They then introduced content into the wild, and what they found was that once there were twenty to fifty citations of a phrase, they saw about a four-times lift in it being correctly referenced.

So the key threshold they identified was roughly twenty to fifty occurrences of that phrase, including consistent contextual usage.

From a search perspective, the immediate reaction is obvious: create twenty to fifty pieces of content, push it into the ecosystem, and effectively “game” the corpus.

But they didn’t see strong deviation indicating spam behavior, as long as the content was legitimate, authentic, and contextually relevant.

Bill Scully (48:34.826)

Mm-hmm.

Bill Hunt (48:52.783)

Another important finding was that in the absence of strong local-language signals, the system defaults to the nearest available language corpus—typically English—to find alternatives or interpret meaning.

So what we’re seeing is a system that now looks more deeply and more broadly at user intent. It pulls from first-party sources—your brand content—and then augments that with consensus from other reputable sources.

And with models like Gemini doing this more frequently, even if only around fifty percent of the time or less, it changes how we need to think about content.

It affects the depth and breadth of content we produce, how it’s structured, and most importantly, how quickly it can be accessed.

This is fundamentally shifting how we approach content strategy.

And I think you used the right word earlier—it has to become more systematic, more engineering-driven thinking.

To wrap up, it would be useful to explore practical things teams can do in the next couple of weeks to test this. For example, take a query prompt in any language and start evaluating how it behaves in different contexts.

Bill Scully (50:37.620)

Before you do that—do you have the name of that research paper?

Bill Hunt (50:44.951)

Yeah, I’ll put it in the show notes.

Bill Scully (50:48.094)

Okay… great. I just wanted to make sure I captured that, because I think that would be pretty interesting.

So yeah—you’re starting to wrap up?

Bill Hunt (51:12.631)

Yeah. So just as a wrap-up, given all of this—now that we have a proper LLM, or Gemini especially on the Google side, looking at our content—how should we be thinking about this?

What’s your take on how we should approach it? Clearly, we can now be loud and proud that web effectiveness is critical.

Bill Scully (51:52.394)
So I think we need to reframe SEO. It is, I mean, you just said not to talk about it again, but I think it’s really web effectiveness.

As far as web effectiveness, what we’re talking about today is web presence management. How do we answer people’s questions when they’re asking? I think we need to stop thinking about SEO budgets as content, and start thinking about how to increase the likelihood that we’re even considered for showing up in AI overviews.

I still like the idea we’ve talked about before of having a VP of answers. Maybe that title needs to change because it’s more than just answers. If we can assign somebody—not necessarily SEO-specific—who is focused on providing a good experience in AI overviews.

Bill Scully (53:11.979)
What are your thoughts?

Bill Hunt (53:16.76)
I’m with you. My next Search Engine Journal article is probably going to be on the argument for the need for a VP of answers—then, like you said, put in parentheses, “or whatever the hell you want to call it.”

I equate it to what you always reference: the Chief Revenue Officer or a revenue or growth manager type role. It’s a person focused on multi-discipline alignment.

I think we really have to take a hard look at what must we be seen for, and what are we actually being seen for. Then take those answers and don’t let the tools do it for you. Sit down in a room with people and ask, what the hell do we do? What would be a loss if we didn’t show up for it?

Bill Scully (53:56.820)
Yeah. Yeah.

Bill Scully (54:51.560)
Right. That’s a really good insight. I think a lot of people don’t realize that people are searching with that level of detail, and without having that detail on your site, you’re not considered.

Bill Hunt (55:07.97)
Exactly. And going back to the peak AI research, one reason so many answers defaulted outside the local market to global answers was query refactoring. It shifted into “what is the best…”

For example, there was a case related to Spain where a large portion of fast fashion companies are based. But none of the Spanish brands showed up because the query was adapted toward global brands, which automatically excluded local ones.

And it’s the same thing. If someone includes business size, I’m not saying you should start hiding content like “we serve one-person companies, two-person companies,” in white text or anything like that.

We’re even seeing people try new tactics, like AI prompt rules—basically telling AI, “this is how I want you to think about us.” There’s a lot of chatter on LinkedIn about people building these and then seeing them influence outputs almost immediately.

And now people are trying to game answers with prompt injection and similar tactics. It’s becoming chaotic.

Bill Scully (57:01.936)
Yeah, it’s the buttons.

Bill Hunt (57:37.687)
That page, and again, this audit I looked at this morning. One thing we deduced—just a really quick example—was that they had a support page, which was one of their top queries.

It had lost five positions in Google and dropped out of the AI overview. When we looked at the page, they just said, “Put more content on it.” But we ran it through Core Web Vitals.

It was getting like an eight. It was painfully slow. So the question became: how can it be that slow if it should only have like five things on it?

What we found was it was calling in a testing tool with an insane amount of lag.

Bill Scully (58:36.210)
It was the site itself that was testing.

Bill Hunt (58:40.110)
The page—this one page—was for how to use their testing tool. It was a web-based testing tool. So when people searched this query, they dropped to number five in organic results and dropped out of the AI overview for their own tool.

When we visited the page, even just loading it took forever for the actual tool to load. We ran it through Core Web Vitals, and there was also no schema on it and no real explanation.

It was just a bunch of offers on what you can buy after using the tool. What we deduced was that the page wasn’t loading fast enough for Google’s systems or even AI to properly evaluate it.

We don’t know the final outcome because they haven’t fixed anything yet—we just saw this today. But based on what we saw, if it’s getting an eight on Core Web Vitals, it’s not good.

Bill Scully (59:50.538)
Wow.

Bill Hunt (59:56.645)
So the point is: I understand it’s your tool, but if other people are showing or talking about your tool, why wouldn’t you want the best experience to be surfaced?

If you’re not showing up where you absolutely should, run Core Web Vitals or ask the system why you’re not showing up.

It may tell you it’s because your site is too slow, or because it couldn’t find information, or something similar.

So do those kinds of checks.

And then the last thing I would say is: try to think like an agent that’s been given a series of commands—“I want to do X,” or “I want a product.”

If you were a bot without human bias, how would you navigate your website to the point of checkout? Can it actually do it? And where would you get frustrated as a human?

That goes back to our friction episode last week—where you get frustrated is probably where a bot would also drop off and look for an alternative. Those are the things I would focus on in the next five to ten days.

Bill Scully (01:01:34.805)
All right, so that was a really good long wrap-up. We covered Gemini 3.5 being out and running much faster, and as a result, we need to make sure our sites are loading fast.

We also know search results are becoming more AI-agent driven.

Bill Scully (01:02:11.904)
And I keep coming back to this—we need someone responsible for answers, and it probably shouldn’t be an SEO. Traditionally, SEO people are measured in ways that don’t align with focusing on AI answers.

Bill Scully (01:02:40.542)
Yeah.

Bill Hunt (01:02:53.026)
And there’s no SEO currently placed at that level in the food chain. It can be someone with an SEO background, that’s fine.

But it has to be someone who can sit—not as a ruler, but more like a growth manager or revenue manager. That’s been your topic forever: the power of a revenue manager to coordinate and motivate teams.

Bill Scully (01:03:15.774)
Yeah.

Bill Scully (01:05:00.010)
Sounds good. Thanks again. And for those listening for the first time, please make sure you subscribe. You can also find Bill, who is constantly writing for Search Engine Land, on Search Engine Land.

Bill Scully (01:05:51.648)
All right. Okay. Thanks, Bill. Take care.

Reach out to us for a consultation.

Need more than training and self-help? Contact us, and we will help you with your International web marketing and technical SEO.

Contact Us

InternationalWebMastery.com
19 Breakwater Ln
Windsor, CT 06095 USA

+1 860.325.0784

Hey AI, Learn About Us

© internationalwebmastery.com

Terms of Service

Privacy Policy

Cookie Policy