From Veterinarian to SEO Expert

Marie's story is wild. She started as a veterinarian and actually treated the pets of Canada's prime minister. As she told me, "I actually saw his daughter's hamster, which was very unfair because you can't do a whole lot with hamsters, but still kind of a cool story."

But in 2008, she got curious about Google as a hobby. She wanted to understand why Google ranked one website over another.

Then in 2012, everything changed. She was pregnant and on bed rest with nothing but time and a laptop. That's when Google launched the Penguin algorithm on April 24th, 2012. Nobody knew what it was about at the time. Marie spent those months chatting in forums and learning everything she could.

One guy came to her and said he'd pay her $300 to remove a manual penalty from Google. She said no at first because she was a veterinarian, not a marketer. But it was a fun challenge, so she did it. She removed his unnatural links penalty, more people came to her, and she wrote a book about the process.

Here's a key lesson: write a book and you become known as the expert. Even today, if you ask ChatGPT about Marie, one of the first things it says is that she removes penalties. All because she wrote the book on it.

Her business grew, she built a team, then after the pandemic she went back to solo consulting. In 2022, Danny Sullivan from Google reached out to her and a few other top SEOs to talk about a new thing they were doing with their algorithms. That turned out to be the helpful content system.

When she saw it was a machine learning system, she realized she needed to understand AI. "I didn't even know at the time if machine learning was the same as AI. It is." Since 2022, she's been learning everything she can about artificial intelligence and now works with a small handful of clients brainstorming on AI agents and the future of search.

The Most Impactful Google Update You've Never Heard Of

I asked Marie what the biggest and most impactful Google update has been throughout her 16 years in search. Her answer surprised me.

It's one that isn't even documented. February 2017.

She had a client at the time and noticed they were being outranked by people who clearly had expertise. All the content that did poorly was what we now call YMYL (Your Money or Your Life) content. This was the early start of Google implementing EEAT into their algorithms.

A year later in August 2018, we had the Medic update. What started in February 2017 got amplified big time. And here's what most people don't realize: this is when Google said they started going AI first.

Since 2017, every update we've seen has been related to Google putting more emphasis on what they can do with AI to determine what sites will satisfy users. Less and less on traditional ranking factors.

Marie's biggest issue with SEO today? "A lot of what we do in the name of SEO is trying to rank for these old systems that don't even exist anymore." When essentially, it's AI that determines what a searcher will find helpful, just like ChatGPT predicts the next word in a sentence.

Google's ranking systems predict what a searcher is going to find helpful. Then they use a massive amount of signals, most of which involve user engagement, to determine if the user actually got satisfied. If so, that fine tunes the system.

User Signals Have Always Mattered, But Now They're Everything

Marie explained something called NavBoost. It's been around since before AI and is essentially a table of user signals. Here's what most people don't realize, and Marie made this crystal clear: "Every single search that you do on Google, Google stores the search query and they store the actions of the users."

Some of the NavBoost signals include clicks (did people click on your result?), long clicks (did they stay or go back?), and longest last click (there's a good chance you satisfied that search).

Google's been storing these signals for a long time. But more recently with AI, they've learned how to actually use those signals effectively. The helpful content system was built on this. They took what they learned from that AI system and incorporated it into the core algorithms in March 2024.

Some people try to manipulate click through rate. Sometimes they succeed short term. But here's the problem: you can't fake user satisfaction. It's one thing to get users to click on your site, but there have to be other signals that really show users were satisfied.

Google also has an AI system called SpamBrain designed to find patterns consistent with manipulation. Every site in Google's algorithms has a spam score. If there are patterns of users using your site that don't line up, that can contribute to your spam score.

Marie shared something fascinating from a 2016 Google presentation called "Life of a Click." They talked about how Google doesn't actually understand web pages (though that's changed now with AI), but they used to really learn from what people clicked on. The clicks of users would tell Google whether they were doing a good job putting things in front of people.

Marie asked me a brilliant question that makes this all click: "If I had to ask you to come up with a rubric of which website is more likely to satisfy the user and you had access to user signals, what's more important, that it has links pointing to it or that there were signals showing that the user was satisfied? Clearly it's the latter, right?"

Exactly.

Google has been collecting data from basically the history of search for the past 10 years. Now with AI, they can take all that information and show the best results. Companies that have existed for 10 years have an advantage because Google has all that behavioral data.

Marie mentioned the DOJ versus Google trial. The main thing about the trial was that Google has such an unfair advantage. Google argues they have a monopoly because they're the best. The judge's decision included that they need to share some of the anonymized user data with competitors.

Google monitors everything. When you're on desktop and do a Google search, they're monitoring what you hover over. They're trying to figure out if the search results page actually satisfied you. Quality raters are rating the top five results, whether they're websites or features like the maps pack.

That's why sometimes you see a maps pack and sometimes you don't. The system has predicted what will be most helpful, then learns from whether people were satisfied.

What Quality Raters Actually Do

The Quality Rater Guidelines used to be a secret. Marie actually bought the domain "leakedqualityraterguidelines.com" to host them when they first leaked. Google sent notices to everyone who published them.

Then around 2017, Google decided to make them public. Their thinking was simple: what's the worst that happens? People produce high quality websites that are more likely to rank. That's a win-win.

Here's how quality raters work. Google contracts 14,000 to 16,000 people. They're given two sets of results: the search results as they exist today, and a proposed change. They rate which side is better based on the guidelines.

The interesting part is that AI is actually the one making those changes. Google does predictions on what to rank, and the weights of the AI decisions are changed by the system itself. The quality raters help fine tune the AI systems to prevent manipulation.

If a quality rater visits your site, it doesn't impact your rankings directly. It's an overall judgment of whether that search result was helpful and whether the algorithm that produced it was the better choice.

Marie checks the guidelines almost every week to see if they updated. When they update, it's really interesting to see what changed because then you know what Google's working towards. In 2022, Google added a ton about real world experience to the guidelines. Sure enough, a couple updates later, pages ranking better were ones that had experience.

Don Anderson wrote a good article a couple years ago on Search Engine Land about how there's a good chance AI will replace a lot of those quality raters. They're finding that AI systems can do a better job than human raters at determining whether new results are better.

EEAT: What It Really Is and How to Build It

There are a lot of misconceptions about EEAT. Marie might have driven some of them because of her 2017 case where a client was outranked by sites with medical expertise. They hired doctors and added medical author bios. Their content improved because they had actual doctors instead of content writers.

But it wasn't about the author bio. It was about the level of expertise that was actually exhibited.

Here's how Marie explained it: "EEAT is not like an algorithmic thing that we can go, you need to have an author bio, you need to have this, you need to have that. It's an overall understanding of all of the signals that exist to indicate whether you're likely to be seen as legitimate, as trustworthy for that query."

EEAT is the world's conception of you. If I asked you where's a good place to get pizza near you, some places jump into your head because of what they're known for. The things you're known for are your EEAT.

Marie told me an incredible story. She hadn't done podcast interviews for a while. Then she did one where they labeled it as "Marie Haynes, SEO and AI expert." The very next day, if you searched for "SEO and AI expert," she was ranked number one. She didn't do any link building. There wasn't even a link. It was that there's an authoritative voice saying she's an expert.

EEAT is what others say about you, not necessarily links. What I'm doing right now with this podcast, interviewing people who understand SEO and marketing, I'm making an association between me and that person. If you write for Search Engine Land or Search Engine Journal, there's knowledge out there that says you're respected enough to be published on these places. That makes a connection between you and the topic.

As Marie put it, "EEAT is mostly what others say about you." A lot of EEAT can be PR, but we need to not fall into the trap of thinking it's links. Links kind of come out of EEAT. When Marie publishes this podcast, even if there's no link to her site, she's put evidence into the world that says I knew enough about this topic to have her on my show.

There's also the element of trustworthiness. Marie is working with a site right now that has struggled after core updates. They have serious reputation issues and problems with returns and refunds. That's a big component of trust. It can be turned around, but it's not an SEO issue. It's an entire business issue.

In the age of AI, there are ways you can game language models right now, but they're getting better. When you ask something of a language model, they're going by all the knowledge in their knowledge base of who is recognized as the best on this topic.

Content can contribute to EEAT too. When Marie writes something, she's not writing it for keywords. She's not writing to potentially rank. She's adding to the body of knowledge that exists on her topic. If you get known as someone who can improve the body of knowledge on your topic, that contributes to EEAT.

How Local Businesses Can Build EEAT

I asked Marie how a local business can build those strong associations and EEAT from a practical standpoint.

Content is one way, but we need to be careful. A lot of businesses are producing as much content as they can to build topical authority. There was a day when that worked really well, but now Google has all that content.

If you publish content that is essentially the same as what everyone else has already published, you're sending signals that your site is not that important. But if you can publish original content, that matters.

Original content might be original research. It doesn't have to be expensive. Just send a survey out to your customers and publish something like "90% of people using pest control have these concerns." It's not difficult to publish something new.

The other way is traditional PR. If search engines didn't exist, what would you do to get your business known? That might involve being on the radio or in the news.

Marie had a client in real estate years ago. There was a tragedy in their town where the school's music room burned down. They did something good for that school, volunteering to collect instrument donations and bring them to the school. They sent out press releases. Not for the links, although they did include a form on their website. They had a ton of people mentioning this realtor in a really good light.

Do good things for your community. Going on podcasts is something that can be good. Anything that puts your name in a good light.

Reviews are really important. Being truly the best business in your town is the biggest thing that matters. That's why SEO is very difficult these days. Marie put it perfectly: "Historically, if I knew how links work, how to do good keyword research, how to structure pages properly, like that's what made sites rank. But now that AI determines what's likely to satisfy users, it's many other signals altogether."

The quality rater guidelines talk about awards too. It's probably not a direct ranking factor, but if your business wins awards, have a page on your website about press or awards you've won.

Marie's Challenge: Write and Publish a Book in One Day

Marie made herself a challenge to write and publish a book on Amazon in one day. This was when Bard just came out. She had a bunch of stuff she'd already written on prompts and how to learn. She used Bard to help her write 10 things you can do to improve your use of Bard.

She used it to figure out how to get published on Amazon. You can actually do it where people can order it as a physical book. She thinks she's sold maybe 10 copies, not many, but she wrote one of the first books on talking to language models.

Just being out there, having written and published a book, you don't necessarily need to do it in one day. Make it a challenge, maybe one month. The temptation will be to use AI. When Marie says she used Bard, she used it to augment her. She'd write something and ask how can I make this shorter? Does this make sense?

None of us really like reading directly AI written content, but using AI to augment your thoughts and improve upon what you've already done is totally acceptable.

If you're trying to build EEAT, write a book on your subject, whatever the topic is that you're trying to be known for. You don't have to sell millions of copies, just have it out there.

AI Agents: The Next Frontier

Marie's number one focus right now is AI agents. She thinks AI agents will give businesses huge advantages and individuals huge advantages. "Those who don't use them, it'll be like trying to run a business without using the internet today. Like you could maybe do it, but you're going to have a lot of difficulty compared to those who have the technology."

There was a McKinsey study showing that 95% of businesses trying to implement AI agents are failing. Marie didn't like the study because it was based on only 52 surveys, but regardless, most people trying to do stuff with agents aren't having tremendous success yet.

The models keep improving. Recently Google released Gemini 2.0 and it's so much improved. A couple days later, Claude 4.5 Opus came out and it's better than Gemini in some benchmarks. Then Grok came out. All these models are improving and improving.

Marie was using Google's new tool called Antigravity when we got on the call. It's like Cursor, where you can code things and use AI to help write the code. Antigravity is a step above.

She's building a tool that helps her analyze which pages did well or poorly after an update. It uses a bunch of agents, which are essentially prompts. Some can use tools like Google search. Each agent looks at content in a different way and can assess if it had original content, fresh content, if it matched user intent.

Other agents analyze their output and come up with patterns. Then other agents give guidelines for creating new stuff on the website in the future. The wild part? "I'm building this in Antigravity without, like I haven't touched a line of code and it's actually working."

She's using Google's Agent Development Kit, which is a Python format for creating agents. She can take those agents and put them on Google's agent marketplace. You might not buy a SaaS tool from her, but you might pay to use her agents.

Google has something called the agent payments protocol that can negotiate payment between agents. If you find her agents useful, you pay a few cents every time you use them. Her agents might have an agreement with your agents that they learn what worked for you and continue to do better.

We're headed towards this ecosystem. Right now, 50% of the time it either crashes or she hits a quota limit. But that will get better. It will continue to improve.

Marie encourages anyone listening to start learning more about what agents are and to play around with trying to build stuff. You don't necessarily need to use Antigravity. If you go to ai-studio.google.com, there's a button on the top left called build. You can just describe what you want to build. You can build an entire website in there and deploy it.

Agents are the way. In a year or two, everyone will be talking about agents. Some people are already building agent workflows with tools like N8N.

Google published a blog post saying every enterprise will soon rely on multiple agents. This is how the world is going to be.

Google recently released a price tracking feature where you can keep an eye on a product. If the price comes down below a certain amount, it'll ask your permission and your agent will go buy it for you. You never went to visit that website.

Marie thinks the web as we know it had to exist so AI could learn. We've been working for Google in populating content. Now we're reaching another age where AI is not just learning stuff but doing stuff for us. We're in the super early phases.

Most of her clients are really interested in building agents, but they're not actually doing anything yet because the tech is not quite there to put something into the public that would represent their brand. But the day is coming.

It's not too late. When Marie gives talks like this, people say they're so lost, so behind, they don't know anything about this stuff. One of the best ways to learn is to take your favorite language model, whether ChatGPT, Gemini, or Claude, and ask it to teach you about how AI agents work. Just have a conversation with it, ask for resources to read.

Google has an incredible group of guides on AI agents, how to use them, how to build them. That's where the future is headed.

The Future of Search: Agents, Glasses, and Beyond

I asked Marie to get very clear and tactical on what the future of search looks like.

The next future of search is agents. AI mode has some agentic features already. In some places, if you're searching for tickets for events, AI mode will do the searching for you. It'll look at Ticketmaster, StubHub, all the places where you might buy resale seats. You have the conversation within AI mode, then you say you want these from Ticketmaster, and AI mode just goes and buys them. You get the email in Gmail. You never went to Ticketmaster's website.

Google announced last week that AI mode can now create a new user interface on the fly. Someone asked a question about Vincent van Gogh, and AI mode changed to be the colors of his paintings. It was a whole different experience than a Google search page.

There will be changes that are hard to wrap our heads around.

The next thing will be browser agents. This is really important for any business that relies on forms or user interaction from within your website. One of the reasons they've created browser agents is to reduce the time a user needs to spend on menial stuff. Nobody likes to fill out a form.

You'll be able to send your agent, which would be Gemini or ChatGPT, to fill out forms for you. Gemini within Chrome is happening. Some areas of the US have it.

We'll have a different way where the search results aren't really going to matter so much.

Marie asked if I saw that Adobe bought SEMrush for 1.9 billion. Everyone's like "Adobe bought SEMrush, that means SEO is not dead." But she thinks they want it for a different reason.

Adobe has all these AI agents. They have a workflow where some agents optimize your website and learn from those optimizations. Some are customer service facing agents that take all your website and brand information and use that to have conversations with your customers. They have a whole suite of agents.

This is the next phase over the next couple years. That's where search is going.

In terms of glasses, Marie finds it hard to grasp that she wouldn't be doing what she's doing now, sitting at a screen. She thinks we're still going to have screens for quite some time. But just like when mobile phones came, we didn't stop using our computers. We just had another screen. Glasses will open up all sorts of new opportunities.

For service businesses, she's really excited about glasses. She doesn't know how yet we take advantage of it, but there will be opportunities. Keep paying attention to what's happening with glasses.

The next thing, and Marie always says "Marie, don't go there" but she goes there anyway, is we will get to the place where we have some type of brain computer interface.

Neuralink, one of Elon's companies, has a small group of people with spinal cord injuries or who can't function who've had a brain implant. They put the first recipient in front of a computer and said "make the cursor move up." He thought about moving up. They were able to read the pattern of neurons that connected to "up" for him.

Now some Neuralink recipients can play Call of Duty with their brains.

At conferences, Marie always asks who would get a brain computer interface, assuming it was safe. The first time she asked, one guy put his hand up. Last time, maybe 20 people in the room said yes.

She absolutely would get it because it will give you such huge advantage. "Every piece of knowledge that's in the world you could just bring to your mind at any time." There's negatives too that would come, she acknowledged.

She doesn't think everybody will need surgery. Lots of people are working on other ways to connect with our devices, like via facial movements. Meta has a wristband that can determine your movements. You look like you're writing your name and it'll write it.

We lived in this little sliver of history where we used plastic keys on a keyboard to communicate our thoughts. That will end at some point. She's looking a decade out, maybe more.

Here's something wild: if you search for "search engine" on Google, you don't see Google. It's always been that way. Google's intent has never been to be a search engine. It's a step along the road.

There's a recent blog post by Demis Hassabis, CEO of DeepMind, the mind behind the AI in Gemini. He talks about Google's goal actually being your personal assistant. He talks about robots and how some of Google's 3D worlds they're creating will be training grounds for robots.

We will at some point have robots in our homes that do many things for us, of which search is a component.

For now, Marie's advice is clear: "We need to really get out of the mindset of search being something we can manipulate with a knowledge of search and rather like search being an algorithm that's putting the best choices in front of people."

Our goal as professionals who advise businesses is to advise them on best business practices. Not SEO, but really being the satisfying choice for users. And measuring that, not looking at keyword rankings, but looking at user time on site, how many people filled out our form, how many conversions we got.

That's far more important than keyword rankings.

Marie's Message: Find Joy in This Time

Marie's final message was powerful. "Work to find joy in the time that we're in. Because there's so much negativity and there's so much fear because we're going through change, right?"

Sundar Pichai said that AI is more profound for civilization than fire or electricity. Marie believes him. We're still in the very early phases.

You'll get much more value from technology in the future if you find the ways to use it and use it to make you better.

Spend time every day with a language model. Your goal is not to go off and make money with it. It's to develop the skill of communicating with AI, because that's where we're headed. We're not going to go back. We're not going to have a future where people decide it was bad and we're not going to do that.

Marie's final thought: "If you can learn to enjoy using AI, life will be a lot easier."

The people who know how to use AI will have such huge advantage over those who do not.

Marie recommends learning about visual search. She saw that one of the top uses of ChatGPT among service professionals was for plumbers who use it to brainstorm ideas.

Google has a Customer Experience Suite where you can do live video chat with AI. For pest control, she thinks it would be very interesting to make some type of app that uses your knowledge base to answer questions for people. There's loads of opportunity there.

My Main Takeaway

This conversation completely shifted how I think about SEO and the future. We're not optimizing for Google anymore. We're optimizing for user satisfaction, and AI is getting better and better at measuring that.

The traditional playbook of keywords, links, and technical SEO still matters, but it's becoming a smaller piece of the puzzle. What matters most is being the best choice for people. Actually satisfying them. Giving them what they need.

And the future is coming faster than most people realize. AI agents will change how we interact with businesses. Glasses will create new opportunities. Brain computer interfaces might sound like science fiction, but they're already here in early forms.

The businesses that win will be the ones that embrace this change, learn to work with AI, and focus on being genuinely helpful to people.

If you want to dive deeper into everything Marie is doing, make sure to check out the full podcast episode. She goes even deeper on so many of these topics, and her insights are absolutely invaluable for anyone trying to understand where search is headed.

Thanks for reading, and if you found this valuable, make sure to subscribe to Marie's newsletter at mariehaynes.com/newsletter and join her community at community.mariehaynes.com. She spends her entire day learning about AI and sharing it, which is an incredible resource for all of us trying to keep up with this rapidly changing landscape.