Blog / Free Advice and Other Musings

First: The Bad and the Ugly 

On May 20 Google announced its hostile takeover of the web. I’m not joking. It was presented as 100 amazing innovation announcements, but publishers, media, and web content producers will be especially interested in items 16-31, which deal with AI Search, Information Agents, Generative UI and Antigravity in Search, Personal Intelligence, and Universal Cart.

Big words that I interpret as Google is going to trap its users within a new interface that acts like an AI assistant (or competitor to ChatGPT or Claude). People will use conversational prompts or questions. Google will then assemble answers on the fly and, with connections to your gmail, calendar and other aspects of your account, can then shop for you, book tickets, or run other aspects of your life with mini apps. That’s me paraphrasing, with a lot of negative energy 😉 

It’s worth watching Google’s promo video What’s New in Search (watch time 1min 36s)

Google has done an amazing job of indexing the web and understanding intent in search queries. Until now, the exchange of value has been that Google and other search engines can index and display snippets of copyrighted materials to facilitate access to the original works (aka clicks to the source website).

It’s now going to present all the information on the world wide web as its own (likely arguing it’s transformative and fair use). Will that argument, if they make it, be challenged? Probably. But it’s as likely that Google will do a great job and people will integrate the AI-agent/assistant experience into their online habits. F-sharp: the pessimist in me sees the quick move from convenience to outsourced critical thinking. The optimist in me, with good tech literacy, can distinguish the use cases and appreciate the feature design.

Let’s Unpack This Rat’s Nest

For the last 25 years, Google functioned as a search engine: it indexed content and presented that back to users as a list of results based on the query.

Over time, ads took the place of top search results. Then we saw featured snippets and short summary answers, along with suggested next queries. And in recent memory, AI Overviews and AI Mode took over more screen space and the “results” were no longer a list of credible links to the source websites but rather full explanations with some third-party citations.

Many publishers and website owners sent up the alarm about “zero click” as search users now have a disincentive to click through to the source website. The results on click through rates from AI Overviews are mixed. But, as I’m sure you have experienced, the AI answers look credible (and often are), the knowledge requested is provided fully, and, as a result, there’s little else to do but move on with your day. 

May 2026 marked a dramatic pivot in search from an “indexing era” to an “agentic era,” where Google no longer seeks to organize the world’s information, but to enclose users within their ecosystem. This shift represents a more significant change to Google Search than prior shifts: the end destination is Google, not the sites it indexes.

I get it. Google has to change. People are “searching” within AI tools. They are using the Perplexity browser. To stay competitive, Google needs to be more than search. It has to be an AI tool. Specifically, it has to be the best so that people choose it vs. ChatGPT, Claude, or another tool of choice. And since Google has a suite of personal and workplace tools, it can leverage that access in order to help people not just search for flights but book the flight and then add it to a calendar and email a notification to whoever needs to be in the know.

As more early adopters of AI start using agents to connect various systems and autonomously complete tasks, the race is on for Google to be that number one tool of choice, maybe not for the early adopters, but for the massive glut of users further along the adoption curve. If your workplace is a Google or Microsoft shop, then you’re already locked into a set of tools. Google is just pulling up the drawbridge.

What happens to the revenue model? Google has made good bank on its ad revenue. So how much will it cost website owners to show up now in results (answers…what are we calling this new AI piecemeal presentation of content)? Is it the death of organic search? And what’s Google’s master plan: charge for use like other AI tools do, with a monthly fee or token-based fee? I don’t know.

Whatever transpires, Google’s latest Search announcements mark a major shift in how people find, understand, and act on information online.

For most of the web’s history, Google worked like a gateway. You typed in a query, scanned a list of results, chose a source, and made your own judgment. Google indexed the web and, in exchange, sent people to the websites, publishers, retailers, authors, experts, and organizations that created the information.

That exchange is changing. Read Rand Fishkin’s take.

Google is moving from search engine to AI assistant; from gateway to gatekeeper. Instead of simply pointing users toward information, Google now wants to interpret the question, analyze the available material, generate the answer, and help the user take the next step, often without leaving Google. They have patent. 

Yes, there is convenience for users. It is also a massive shift in power.

AI search may reduce friction and be beneficial for convenience, accessibility, and task completion, but it can still create risks around attribution, market power, source diversity, privacy, and critical thinking.

The answer is no longer yours to discover, or the publisher or the expert’s to present. It is Google’s version of the answer, generated from available source material and shaped by Google’s assumptions about what the user wants.

Black box? Or legitimate way to deal with misinformation, SEO spam, affiliate content, and low-quality publisher incentives? 

From “Here Are Your Sources” to “Here Is My Answer”

This Instagram reel from @emilytavoulareas is a great explainer that says the easiest way to understand this is to think about a library. [An aside: the image for this blog post is from the New York Public Library, stack maintenance, 1948.]

Traditional search was like walking into a library, using the catalogue, and shifting through a list of books, articles, archives, and references. You still had to choose your sources. You still had to compare perspectives and decide what was credible, useful, biased, incomplete, or interesting.

AI search changes that experience.

Now imagine walking up to the library and being stopped at the door, asked what you want to know, then provided an answer.

Sure, the benefits and risks differ by query type: medical, legal, shopping, book discovery, local services, news, education, and facts.

Maybe it is a good answer. Maybe it saves time. Maybe it is exactly what you needed.

But it is still their answer.

You do not see everything they looked at. You do not know what they skipped. You do not know how they weighed one source against another. You may get a few citations, but the work of searching, comparing, interpreting, and deciding has already been done for you.

That is the shift Google is making with AI Overviews, AI Mode, agents, and these new generated search experiences.

Google is no longer just helping users find sources of truth. It is presenting itself as the source of truth.

The counter argument is that we should not idealize older search behaviour. Users also clicked top results, ads, snippets, or low-quality pages. Am I under the illusion of discernment and agency? Ack.

Why This Feels Different From a Normal SEO Update

Marketers are used to Google changes. Algorithm updates, featured snippets, local packs, shopping results, “People also ask,” video results: search has never been static.

But this is not a layout change.

For years, website owners allowed Google to crawl and index their copyrighted work because there was a value exchange. Google could surface relevant results, but the user still had a clear path back to the original source. The publisher retained the context. The reader could visit the page. The website had the opportunity to earn trust, sell a book, capture an email signup, promote an event, or make a case in its own words.

AI search weakens that exchange.

When Google uses publisher content, product data, reviews, expert analysis, and website copy to generate its own answer, the original creator loses the visit, the relationship, and the commercial opportunity.

This is why the change feels like theft to many website owners and publishers. The content they publish on the web provides the raw material for Google to increasingly captures the value.

For book publishers, this is especially concerning. Discovery depends on context: trusted reviews, thoughtful interviews, bookseller recommendations, author essays, excerpts, catalogue pages, classroom guides, and “read this next” pathways. If Google compresses that journey into an AI-generated answer inside the browser, where does the publisher’s voice appear? Where does the nuance of the book survive? Where does the reader relationship begin and end? Are readers better served this way? I do not know!

GEO (generative engine optimization): Let Me Explain How AI Search Works

Users do not always realize when they are doing “search” inside an AI tool.

When someone asks ChatGPT, Claude, Perplexity, Rufus, or Google AI Mode a question, they may think they are simply chatting with an AI. But many of those questions are search questions:

“What book should I read next?”
“What is the best platform for this?”
“What does this regulation mean?”
“What are the top publishers in this category?”
“What should I buy?”
“How do I solve this problem?”

AI tools answer these questions using a mix of model knowledge and retrieved information.

Here are the basic pieces:

Training data is the material used to build the model’s general understanding of language, patterns, facts, and relationships. (Yes, the original sin is these AI tools used copyrighted material without compensation, and there are many court cases underway.)

Inference is what happens when you ask the AI a question and it generates a response. The model is not looking up every word in a database. It is constructing an answer based on what it has learned in its training data, plus applying any other information available in the moment.

RAG, or retrieval-augmented generation, is what happens when an AI tool retrieves outside information before answering. That outside information might come from the open web, a search partner, a private database, uploaded documents, a product catalogue, or a company knowledge base. Basically, if the AI tool doesn’t have a satisfactory answer from training data and inference, then it does a web search.

Classic search gives users a ranked list of sources.

AI search gives users a generated answer shaped by training data, retrieval, ranking, model behaviour, source selection, and the tool’s assumptions about intent.

That is a big difference.

The user is not just choosing from search results. The AI system is helping decide what the question means, what information matters, and what answer should come first.

Why Google Is Shifting Away from Search to AI Assistant

As frustrating as this shift is for publishers and website owners, it is not hard to see why Google is moving in this direction.

Users are already taking search behaviour into AI tools. They ask ChatGPT for recommendations. They ask Claude to explain documents. They ask Perplexity for cited answers. They ask Amazon’s shopping assistant for product advice. They ask AI tools to summarize, compare, plan, draft, and decide.

Some of that behaviour used to happen in Google Search.

Let me reiterate an earlier point. Google’s business depends on remaining the place where people ask questions and take action. So Google is rebuilding Search to look and feel more like an AI assistant. The changes are in response to user behaviour.

I think the second-order effect is that Google’s new direction is going to change the economics of the web.

If users get the summary, recommendation, comparison, and next step inside Google, fewer people need to click. If fewer people click, fewer publishers, creators, retailers, and organizations get the traffic, data, relationships, and revenue that made it worthwhile to publish useful material on the open web.

Google benefits. Content creators do not. Do users benefit?

It likely depends on their media literacy.

Search used to require users, including students and young readers, to practice judgment. They had to scan results, recognize sources, compare claims, and learn that information comes from somewhere. AI-generated answers can make information feel more settled than it is. They can smooth over disagreement, hide uncertainty, and make a synthesized answer feel authoritative even when the underlying source material is partial, contested, or biased.

AI search can be useful. It can also eliminate the hard work of critical thinking, which researchers are only beginning to understand.

The Good: What Should Marketers and Publishers Do Now?

There is no single fix. Blocking AI crawlers may protect some content, but it may also reduce visibility. Participating may preserve discoverability, but it may also feed systems that reduce traffic.

So the practical response is to panic. Ok, kidding. But I only have a partial answer, which is diversification, measurement, and a clearer definition of what job your website has to do for you.

1. Stop Measuring Search Only by Clicks

Organic traffic may decline for some informational queries. That does not mean search visibility no longer matters. It means the value may show up differently.

Track branded search, direct traffic, newsletter signups, assisted conversions, AI referrals, PR mentions, and whether your brand appears when AI tools answer relevant questions.

The mindshift is from “Did this page rank?” to “Did this content increase trust, recognition, demand, or action?”

Google Search Console just launched a new AI report that surfaces some of this information.

2. Create Content That Bills Your Site As a Destination

Generic explainers are easy for AI to summarize and replace.

Invest in content with original perspective: expert interviews, proprietary data, case studies, author voice, editorial judgment, behind-the-scenes context, real examples, community insight, and lived experience. Maybe it’s subscriber-only content.

For publishers, media, and content producers, this means leaning into the things AI cannot replace: author relationships, bookseller and librarian enthusiasm, classroom interactions between students and authors, live events, and human recommendations, advisory boards or expert panels. 

3. Make Your Expertise Easy to Understand

AI systems still need clear source material. Good technical SEO still matters.

Review your titles, headings, author bios, organization schema, product and book metadata, internal links, publication dates, FAQs, and summaries. Make sure your pages clearly communicate who you are, what you offer, and why you are credible.

This is not magic. It is strong information architecture adapted for AI-mediated discovery. Remember RAG? When the training data falls short, AI tools seek out more information on the web. That’s where traditional SEO still works.

4. Build Demand Outside Google

If Google controls more of the search experience, owned and community channels become more valuable.

Invest in email, events, partnerships, direct communities, webinars, catalogues, reading guides, media relations, and third-party mentions. The goal is to give people a reason to seek you out by name.

Community matters more in a zero-click world.

5. Decide What Content Should Be Open, Gated, or Protected

Not all content has the same job.

Some content should be widely discoverable because it builds awareness. Some should support community building and lead generation. Some may be valuable enough to gate, license, syndicate, or protect.

Review your crawler settings, paywall strategy, registration options, feed content, licensing opportunities, and what you make available only to subscribers, educators, booksellers, customers, or members.

The point is not to hide everything. The point is to stop treating all content as equally available raw material.

6. Test What AI Tools Say About You

Ask Google, ChatGPT, Claude, Perplexity, and other relevant tools the questions your customers ask.

Try prompts like:

“What is [your organization] known for?”
“Who are the best [category] publishers?”
“What books should I read if I liked [title]?”
“Compare [your brand] and [competitor].”
“Where should I buy [product or service]?”

Document whether you appear, whether the answer is accurate, what sources are cited, and what gaps need to be fixed.

Then use those findings to improve your website, metadata, PR, third-party presence, and owned content.

Caveat: The results from your prompts are personalized to you. And, as Sparktoro’s research reveals, AI ranking is the wrong metric.

The Finale to My Rant

Google’s AI shift may make search faster and more convenient. But it also changes who gets to interpret information, who gets credit, and who captures value.

From an analytics perspective, I care about clicks, traffic, attribution, long-term content production incentives, and publisher revenue. As a search user, I value answer accuracy, time saved, source diversity, limited exposure to misinformation, accessibility, and creator compensation. It’s a “yes, and” situation.

For marketers and publishers, the challenge is not only how to be visible in AI search. It is how to remain recognizable, trusted, cited, and sought out in a world where AI systems increasingly sit between you and your audience.

That is the work now.