THE AI ADVANTAGE - Issue No. 9

Claude Found Us 9,000 Searches. None of Them Were Customers.

What a week of real SEO research taught me about the one thing AI still cannot do for you

THIS WEEK IN AI

Remember last week when I told you about Cowork and the Ubersuggest MCP and how I was suddenly able to have Claude actually log into things and do the work for me? Well this week I put all of that into a real client project, and friend, I have stories.

Specifically, three of them. Three moments where Claude was about to lead a perfectly nice client down a perfectly wrong path, and I caught it just in time. (One of them I almost missed. We'll get there.)

So that is what this whole issue is about. The honest, sometimes embarrassing, mostly funny experience of doing a real sixteen-page SEO research project alongside an AI that is very fast, very capable, and very willing to be confidently wrong if you let it.

Grab your coffee, tea, or mushroom chai. Let's go.

OPINION

AI IS A BRILLIANT JUNIOR EMPLOYEE WITH NO COMMON SENSE (AND THAT IS ACTUALLY FINE)

I spent two days on one project. A full SEO and AEO research run for a client who owns an aesthetics clinic. Sixteen pages, fully researched, with a final keyword strategy delivered at the end.

It took me all day and then some. The kind of all day where you look up and it is somehow 6 pm, and you think, "crap, I don't have a plan for dinner yet!"

In fairness, a lot of that was quality control. I was testing the whole way through, watching where Claude got things right and where it got things directionally wrong, because we are slowly building this into a skill we will share. Once that skill is finished, this exact same process should take about thirty minutes of actual research plus another fifteen for human review. Forty-five minutes total. For a project that used to a full day of my time.

Here is the part I want you to sit with for a second. The pricing to our clients does not change. What changes is how much of my day this kind of work costs me. Which means I can take on more clients without burning out, deliver faster, or honestly just take a long lunch outside with the chickens. The time decreases, the profit margin grows, and the client still gets the same expert-built deliverable they were always going to get.

That is the cherry on top of a delicious caramel sunday. Not "AI replaced my work." More like, "AI gave me my evenings back."

But (and this is a real but) only if you are actually paying attention. Which brings us to the part where Claude almost helped me ruin a perfectly good aesthetics clinic.

HOW TO USE AI THE RIGHT WAY

THREE TIMES THIS WEEK CLAUDE WAS ABOUT TO LEAD MY CLIENT OFF A CLIFF

I started the SEO research the way I start every project. I gave Claude every detail I could think of. The business name, the city, the address, the type of services, who the customers are, what kind of storefront she has. The plan and breakdown of her website. I told Claude she is a local aesthetics clinic with a brick and mortar location and her customers live within driving distance of her front door.

Then I let it cook (as the youngsters say).

Mistake One: Claude was scoping globally and didn't tell me

The first round of keyword data came back and the numbers were huge. Like, "we are going to make her rich" huge. I was already half-celebrating when something in my gut paused.

You're telling me this many people in a little town are searching for lip filler? If so, we should be investing in lipstick!

So I asked. Where exactly did you pull this data from? And Claude said, "Oh yeah, you're right, I searched globally, not the US, and not her town specifically."

(Notice that Claude did not flag this on its own. It just ran the search the way it wanted to and handed me numbers that looked great on the surface. The geography question had to come from me.)

Now here is where I want to be careful, because I know a lot of you reading this run online businesses, and global search volume actually matters to you. If you sell a book, a course, a virtual workshop, anything someone in Australia can buy at 2am in their pajamas, then yes, the global number is the one you want to be looking at. 

But if your customers have to physically drive to you, walk into your storefront, or hire you for a service in a specific city, global volume is a vanity number. It looks great. It tells you almost nothing useful. You need to scope your research to the country, region, or specific city where your actual customers live.

The lesson here is not "global data is bad." It is "know which one your LLM is pulling for you, and make sure it matches the kind of business you are actually running."

I asked Claude to re-run the research scoped to the US, and then to her specific city. The numbers shrank dramatically (obviously). Most of the top contenders held their position, which was reassuring, but the actual opportunity sizing was now real instead of imaginary.

Mistake Two: It pulled the wrong difficulty score

The next one was quieter, but if I had not caught it, it would have been ugly.

When Claude scored each keyword for difficulty, it pulled paid difficulty. Paid difficulty tells you how competitive it would be to run Google Ads against that keyword. Which is great. If you are running Google Ads.

She is not running Google Ads. I needed organic difficulty. The score that tells you how hard it would be to actually rank a page in regular old search results, with no money behind bumping it to the top.

While the numbers CAN correlate, they don't always. If I had built her whole content plan around the paid difficulty numbers, we may have missed some keywords we could actually rank for! 

Mistake Three: The one I almost missed

This is a big important one so lean in!

When I got to her about page, Claude got really excited. The phrase was nurse practitioner aesthetic. Almost 9,000 searches a month. A perfect match, since she is a nurse practitioner who works in aesthetics. Claude told me this was the winning keyword for her about page. An untapped gem.

But something in my brain said wait a minute, this doesn't seem right. Are customers REALLY searching that? I certainly was not searching that the last time I was looking for beauty services. (When I am hunting for a good facial, "nurse practitioner aesthetic" is not what I type into Google. I am typing things like "best skin laser therapy in Atlanta.")

So I asked Claude to look at the actual search intent behind that keyword. Who is typing this phrase into Google, and what are they actually looking for?

The answer came back as I suspected and I almost spit out my chai. (Ok, ok, I am being a little dramatic.)

The people searching "nurse practitioner aesthetic" were almost all aspiring nurses. They wanted to know how to become a nurse practitioner in the aesthetics space. They were asking about salaries, training programs, schooling costs, how long the certification takes. They were career-shopping. They were not shopping for Botox and laser treatments.

Zero of them were potential clients.

If I had taken Claude's word for it, that about page would have started ranking for the wrong audience. Her traffic numbers would have looked amazing. Her booking calendar would have stayed empty. She would have been the queen of search results for people who would never become her customers.

We aren't just striving for traffic. We want traffic that leads to conversions. And Claude totally missed it before I questioned it.

REAL TALK

AI CANNOT THINK CRITICALLY FOR YOU

In all three of those moments, Claude did exactly what I asked it to do. The tools all worked. The MCP worked. Cowork worked. Nothing was technically broken.

What was missing was the part of the work a senior strategist brings to the table. The little voice that says wait, who is actually searching this, and why? The instinct to question a number that looks too good. The judgment to realize that "nurse practitioner" plus "aesthetic" means one thing to a 24-year-old at nursing school and a completely different thing to a 38-year-old looking to book a chemical peel on her lunch break.

That voice did not come from Claude. It came from my learning the ins and outs of SEO over the last couple of years and doing this work many times the hard way.

Think of AI like a brilliant junior employee on day one. Totally eager, fast, and already better than most people at the mechanical parts of the job. But still a junior. Still needs someone senior reviewing the work and asking the questions that come from experience.

This is also exactly why we are not rushing the SEO skill we are building. It is real. It is mostly done. But we are still testing, still doing the quality control, and finding the small places where Claude needs more direction and guardrails. (Like, you know, please ask if this client is local before you go pull global search volume.) When we put that skill in the shop, we want every person who uses it to land on the right side of all three of those mistakes automatically.

That is what people are actually paying for when they buy a skill from someone they trust. Not the prompt but the expertise and strategy baked into it. So when you are out there shopping for AI skills, look for the people who are actually using the thing they are selling. The shiny prompt that just got slapped together by someone who has never run it on a real client is not the one you want. (And there are a lot of those right now.)

YOU CAN TRY THIS TOO

LOOK AT YOUR OWN WEBSITE LIKE AN SEO DETECTIVE

Even if you do not have a tool like Ubersuggest yet, here is one move you can make this weekend.

Open Google Search Console. (It is free. If you do not have it set up for your site, that is officially homework.) Look at the queries report and find the pages getting impressions but very few clicks. (If you are just setting it up, you will need to give it a little time to collect some data for you, so don't wait! While you are at it, set up your Google Analytics, Google Business Profile, and Bing Places for Business!) 

The data will give you valuable insight and tell you something specific. People are seeing your page in search results, reading your title and meta description, and deciding no thank you. The wording is not signaling that your page is the answer they came for.

We just caught this on our blog post about one-page websites. It had 1,300 impressions and only two clicks. The post itself is great, but the title and meta description must not make people feel like the post would solve their problem.

The good news with this data is that it's fixable! We can adjust the title, rewrite the meta description, and republish it. And hopefully, watch the click rate increase. One quiet little blog post can turn into a real source of traffic in a week, without writing a single new word of content.

(If you are avoiding writing blogs because it's too time-consuming, but you know they could help you get found, check out our Blog Writing Skill and Schema Markup Creator! It's created to write unique blog posts that help AI recommend your business. Members of the AI CLUB get an exclusive discount.)

FROM OUR DESK

THE QUESTION WE WANT YOU TO SIT WITH THIS WEEK

The most important thing AI has taught me this year is not how to move faster. It is how to ask better questions.

Every one of the three near-misses above had the same root cause. Claude moved fast. It did not pause to ask itself the question a human would have asked. So the responsibility to ask it landed right back on me.

That is not AI failing. It is just the actual reality of working with it. The people who get the most out of these tools are not the people who let AI think for them. They are the people who use AI to do the work faster, and bring all of their hard-earned human experience to partner with it.

So here is the question we want to leave you with this week.

What is something you do over and over, every week or every month, that AI could help you do in a fraction of the time that would lead to more profit in your business?

Once you find it, your pricing does not have to change. Just because you can do it faster does not mean you should charge less. Think about it this way. Should a logo designer be penalized for being able to nail your vision on the first try in two hours? No. If anything, she could charge more, because she is getting you results so much faster.

Faster work means more profit, more clients, quicker turnaround times, or honestly just a little more breathing room in your week. Whichever one of those you need most right now. This brings in a lot more opportunities for you!

See you back here next Sunday.

XO Krissy + Claire

BONUS | For AI Club Members

This week's featured deal in the AI shop — 30% off the Blog Writing Skill + Schema Markup Creator. Use it the next time you want to write a blog post that actually shows up in search results (and gets clicked on).

The Profit Margin Multiplier Prompt: This prompt is going to help you identify the 20% effort in your business that makes 80% of your results! If you haven't done this exercise it's gold! (and you should do it quarterly, not just once.) It will help you identify where AI can assist you in making more profit in your business!