It's Been 100 Days Since I Launched AI Girlfriend Coach. Here's What Happened.

I started it as an affiliate site. I'm not going to pretend otherwise.

I started writing my first reviews on July 27, and on August 3 I launched AI Girlfriend Coach with seven of them. The goal was simple: build an affiliate website that makes money. That was the plan.

The other half of the plan was the reviews themselves. A lot of what I was finding online about AI girlfriend apps read like it was written from a feature list. Same structure, same praise, no sign anyone had actually paid and used the thing for more than an afternoon. I wanted mine to feel more genuine than that.

So I decided early that I'd pay for the subscriptions myself instead of taking free access, and show it. The dates, the screenshots, the receipts, the name that shows up on my bank statement. If I was going to ask anyone to trust a review, I wanted the proof sitting right there next to it.

I had to learn how to review these apps

Here's the honest part. When I started, I didn't have much experience with AI girlfriend apps. I knew how to write. I knew how to research. I didn't know these products.

I was also worried about getting things wrong. I was especially cautious with my first reviews, Candy AI, Nomi, Kindroid and Replika, and careful about every claim I made in them.

My first reviews followed a template, the same sections in the same order for every app. It looked tidy, but it didn't fit how I was actually using them. What happened to me on day three of an app didn't belong in a "features" box. It belonged on a date.

That's how the living reviews came about. One review per app, a verdict on top that always reflects everything I've seen, and a dated timeline below it that keeps growing. When an app changes, the review changes. When I get something wrong, I fix it, and I made my share of mistakes in those early reviews.

I knew how to write and how to research. I didn't know these products. The apps taught me that part.

The experiments I'm proudest of

None of these started as a big plan. They started with something small I noticed, or wondered about, while testing.

• The 7-day memory test (August 24). This one came from picturing a normal conversation. You tell someone about a trip you have coming up, and a week later they bring it up, because they remembered. Would an AI girlfriend do that? So I tell each app about the same trip in one natural message, bury it under a week of small talk, then ask one vague question. Some apps remembered every detail. Some invented a trip I never took.
• The image test (August 28). This one started with HeyGF. Its photos looked so much more realistic than what I was getting elsewhere that I wanted to know if I was imagining it. So I sent every app the same request and put the results side by side, so you can see the difference instead of reading about it.
• The Reddit count (September 24). I counted which AI girlfriends Reddit actually recommends, instead of guessing.

I'm proud of every one of them. They're also still running: apps keep joining the memory test as I review them.

The articles I couldn't have written without testing

As I continued testing AI girlfriend apps, I found myself writing about things beyond individual reviews. These weren't articles I'd planned when I launched AI Girlfriend Coach. They came from things I experienced, questions I started asking, or problems I ran into along the way.

I couldn't have written these articles when I started. I simply didn't have the experience yet.

That's something I hadn't anticipated. I started AI Girlfriend Coach thinking I'd mostly be reviewing apps, but actually using them kept giving me new things to write about.

Why I paid for a survey of 2,150 people

Testing taught me how these apps work. What features land, what users probably won't like, and the things people need to see before they pay: pricing, images, memory, conversation.

But after a lot of tests, something started bugging me. All of this testing was for users, and I still didn't really know what users think.

Using these apps is my job. I test them because I review them. The people who use them for real have their own reasons, and I couldn't find enough data on what those reasons are. It's easy to search, but there isn't much there.

So on September 1, I published our AI relationship survey: 2,150 U.S. adults asked about AI relationships, with 1,709 responses kept after a quality screen I explained here. It cost money, and it's a completely different thing from reviewing apps. I'd already spent money on subscriptions, though, and I figured research was a better investment than most.

I'll be honest about the other reason too. I'd be lying if I said I didn't hope it would get AI Girlfriend Coach some attention. But that wasn't the main point. Not many studies look at AI girlfriends or AI companions, and I thought the data could help people, even developers. Most of all, I wanted to understand the reasons behind the use, so I could test better and review better.

I understood the products. I wanted to understand the people using them.

Pitching journalists: cited, but not clicked

After the survey went live, I pitched the findings to journalists myself. Some replied. Most didn't, to be honest.

But a number of articles did end up citing the study, and seeing my work referenced somewhere else felt good. I won't pretend it didn't.

One of them is The Face They See's guide for women whose boyfriend has an AI girlfriend. It used our numbers on whether AI romance counts as cheating, and got them right.

Did it change the site? For a while, a little. What it really did was set AI Girlfriend Coach apart. When I looked around, I found that only a few review sites in this niche actually test the apps, only a few show screenshots and receipts, and fewer still spend their own money. The survey added one more thing on top of that.

Traffic, though? Not really. The coverage didn't bring a flood of visitors. I also shared my thoughts on Medium, Substack and Indie Hackers, and I'd say the same about those. They built credibility. They didn't build traffic.

Credibility and traffic turned out to be two different things.

The hardest part: slow traffic, and running out of ideas

The most frustrating part has been the slow traffic growth.

I've done the things I believe in. I've written reviews as genuine as I can make them. I show what I spend, the receipts, the screenshots, even my chats with the AI girlfriends themselves. I've put out industry research. I'm about as transparent as a review site can be, and I'd be lying if I said the slow growth didn't frustrate me.

Here's my Google Search Console for my first 100 days, August 3 to November 10. Over that stretch, Google sent me 572 clicks from about 15,900 impressions.

There was some encouraging visibility in August. Then it dropped hard in early September, and since then, getting consistent traffic from Google has been one of my biggest challenges.

It's frustrating, after all the time I've put into testing apps, publishing reviews and doing original research. But I'm building AI Girlfriend Coach for the long term, and 100 days is still early.

Right now I rank for my own brand name, AI Girlfriend Coach. For the keywords I actually want, Google does show me, just not where anyone looks. In 100 days, "best AI girlfriend app" showed my site 587 times, at an average position of 41, around page five. That got me two clicks.

Add up every search containing "AI girlfriend app" and it comes to about 2,000 impressions and 11 clicks. My best single day was October 30, at an average position of 22. That's page three. For most other keywords, I don't rank at all.

The second hard part surprised me more: running out of ideas. Not ideas in general, ideas worth publishing. Some of my best days were when an idea showed up on its own, like the memory test, the image comparison or the Reddit count. Those moments made me proud. But there are other days when I sit there trying to think of the next post or guide, and nothing I come up with is good enough to put my name on.

What made it worth it: developers started writing to me

If you'd asked me on day one what my biggest accomplishment would be by day 100, I'd have said traffic or affiliate money. It turned out to be neither.

Part of it is just the work itself. Testing these apps day by day, writing reviews I believe in, and watching the roster grow, one living review at a time. Seeing it all live makes the effort feel paid off.

But the moments that really stuck with me were the emails. Developers and founders of AI girlfriend apps, and sometimes the people doing their marketing, found my contact and wrote to me. Some asked me to test their app. Some said my reviews were among the most genuine and thorough they'd seen. Some had come across the 7-day memory test and wanted their app put through it too. That's how NaughtyChats, EroLove and Hayati came to me, and I still paid for each of them myself.

What I like even more is when it goes both ways. I've traded emails with founders where I pointed out something that needed fixing, and later got a reply like, "Hey, we fixed this, it's live, can you check it?" On Affiny, for example, signing in with Google kept sending me back to the homepage. I reported it, and support wrote back that I'd found a real bug, and that it was fixed.

And sometimes the useful part is for readers. On NaughtyChats, the Google sign-in screen said I was signing in to a long supabase.co address instead of the app's name. Supabase is a backend a lot of new apps are built on, so I didn't treat it as a red flag, but it looks suspicious if you don't know that, and it's worth knowing before you sign in. That's the kind of thing I want a review to catch, for the people using the app and for the people building it.

I'm not earning from it yet. But the people building these apps are reading the reviews, and sometimes changing things because of them.

That doesn't mean every review I write is right. What makes those emails meaningful is that they come from people whose products I've looked at critically.

The numbers on day 100

Here's where things stood on November 10, day 100. The live numbers are always on my updates page.

One thing worth clarifying: I haven't placed any affiliate links on AI Girlfriend Coach yet.

So the $0 in affiliate earnings isn't the result of people clicking my recommendations and not converting. I simply haven't started monetizing the website through affiliate links.

That was always part of the original plan, and I still intend to introduce them. But so far, I've been concentrating on testing apps, publishing reviews, and building the site.

When I eventually add affiliate links, I don't intend to change how I test or recommend products.

The spend is every subscription charge, from my own expense sheet. Character.AI is the one app I reviewed but couldn't pay for: it declined my cards.

What I'd do differently

I tried to think of something I'd change, and honestly, I couldn't find much.

I made mistakes. I had lapses. Some errors made it onto the site before I caught them. But all of that is how AI Girlfriend Coach became what it is now, and I don't think I'd have the testing approach I have today without them.

If there's one thing, it's that I'd start earlier. It took me a while to find the courage to launch this website at all. And I'd probably start the experiments sooner, because they turned out to be the most useful thing I do.

I'm not satisfied with the results yet. The traffic isn't where I want it, and neither is the money. But I'm happy with the progress, and I don't regret the work.

What's next for AI Girlfriend Coach

For the next 100 days, I want to focus on working with AI girlfriend developers and testing new platforms.

More developers are reaching out and asking me to review their apps, and some of those apps I haven't had the chance to try yet. I'm trying to take on as many of those requests as I reasonably can. These are apps that are already live, and if they're new and nobody's looking at them, a review does two things at once: it introduces them to people who've never heard of them, and it catches problems the developers may have missed.

Let me be clear about what "working with developers" means, though. It doesn't mean paid reviews. I pay for the subscriptions I use in my reviews rather than accepting complimentary access, with Character.AI the one exception, because it declined my cards. And every review still documents what actually happens when I use the app, good, bad or somewhere in between. A developer asking me to test their app doesn't get a better verdict for asking.

I also want to go back to apps I've already tested. I started with Candy AI, my first paid app, and resubscribed in October. I've paused those follow-ups for now because of how many new requests are coming in, but I don't want older reviews, or the AI girlfriend apps I recommend, to turn into snapshots of products that have since changed.

Then there's voice and video. So far I've spent most of my time on conversations, images and memory. I think voice and video are where these apps need the most work. Video costs a lot more to generate than an image, and I want to know whether the results justify it. And while some AI tools already sound remarkably real, a lot of the AI girlfriend apps I've tried still sound too robotic to me.

As for more original research, I'm keeping my options open. I don't want to run another survey just for the sake of having one. I'd rather let the next idea show up the way the memory test and the image comparison did.

One hundred days later

When I launched AI Girlfriend Coach, I wanted to build a profitable affiliate website, with reviews that felt more genuine than most of what I was finding online.

One hundred days later, it isn't profitable. There are no affiliate links on it yet, the traffic is still modest, and there's a lot of work ahead.

But I've tested 27 apps, spent my own money on every subscription I could, run experiments, paid for original research, and had conversations with developers who used my feedback to fix their products.

None of that was part of some grand plan when I published my first review. Most of it happened because I got curious, noticed something worth looking into, or just wanted to understand an app better.

The site hasn't reached the numbers I hoped for. I'm happy with what I've built anyway, and I'm looking forward to the next 100 days.