From release to organic traction: how Google and ChatGPT took over from Reddit
This dev log is about Inselnova: one island, sea wars and alliance politics in a world that never resets. Play free →
TL;DR: Inselnova launched in February, and for the first four months nearly every new player came from something I did: messaging people, posting on Reddit, a friend sharing it in their Discord. In June I built pages comparing the game to others, and on 30 June I finally connected Google Search Console. By August, Google and ChatGPT were sending most new players every week without me pushing anything. ChatGPT went from one signup a month to 52, and those players stay longer than the ones Reddit sends. That’s enough traffic to let me work on the game instead of marketing it.
"Hi! This game was suggested by chatgpt, when I asked him to find me strategy or simulation games (on the web) that are available for free. (no P2W)"
Direct messages first, then Reddit
The first players came from me messaging people directly, and that got the game to around ten new players a week through February and March. Once it was somewhere I was happy to show, I posted on Reddit. The week of 30 March brought 73 new players, 49 of them from Reddit. I posted a couple more times after that. None of it was a campaign. I wanted people in the game so I could see what happened: did they get past the first screen, did they come back.
The biggest bump didn’t come from me at all. A friend posted the game in their Discord, and the week of 13 April brought 322 new players. Most of them show up with no referrer, because Discord strips it, but the players told me where they’d come from. By May it was back to 10 or 20 a week.
Comparison pages, which I didn’t think would still work
Around then I asked a friend, Chris, for help. He works in marketing, and he’s the person I go to for marketing advice. He also makes Bangkok Legends. A lot of what he tells me is the obvious thing that’s been in front of me the whole time, which I was too close to see. He told me what I was doing wrong and pointed me at a set of SEO skills, including the Squirrel site scanner, which I built into my own setup so the whole process is repeatable. Every month he sends me a marketing report too. It gives me some direction, and I use it to fine-tune my SEO and marketing mentor AI skills.
I also listed the game on game directories for the backlinks: itch.io, IGDB, IndieDB, browsergamestop and others. New game indexing sites pop up every week, so there’s always somewhere else to add it.
Then, between 1 and 10 June, I built pages comparing Inselnova to the games people already search for: Grepolis, Travian, Ikariam, Tribal Wars, OGame, and a round-up of the best browser strategy games. I was dubious. It felt like an old-school way of doing things and I wasn’t sure it still worked. I also didn’t connect Google Search Console until 30 June, which I should have done on day one. From the moment it was connected you can see Google working through the site, slowly. It picked the pages up within days, but it took about two months to index most of them. The 110 it hasn’t indexed are pages from the live world, the governor profiles and the like, which I’d deliberately kept out of the index. I’m adding those in now.
Asking every AI about my own game
In May I tried something I’m still not sure worked. Once a week for a few weeks I opened ChatGPT, Claude, Gemini, DeepSeek and a few others, logged out, and asked them for browser strategy games, leading them towards mine. Once one found Inselnova, I kept going: compare it to another game, why should I pick it, who built it. My thinking was that if they’re searching for the game, they’re seeing people ask about it, and some of that might end up in what they learn from. Half of it was ego searching, seeing what they said about the game and about me.
Only one of them sends anyone. ChatGPT has sent 206 visitors. Every other assistant combined has sent 9.
What I can’t do is reproduce it. When a player tells me what they asked ChatGPT, I type the same thing and usually don’t get Inselnova back. People are on different models, and ChatGPT personalises its answers to the person asking, so I don’t think there’s a way to see what they saw. Google is easier. The main thing that changes a Google result is where you’re searching from.
August, when Google and ChatGPT took over
ChatGPT sent 1 new player in May, 5 in June, 6 in July, then 52 in August and 33 in the first half of September. Google went from 24 a month to 77. Reddit, which brought 139 players in April, sent 19 in August and 8 so far this month. From the week of 10 August, Google and ChatGPT between them have sent most of every week’s new players. That’s about ten weeks after the comparison pages went up.
The Search Console numbers show the same turn. Clicks have held around 20 a day since mid-August, and impressions have climbed from there to nearly 200 a day.
ChatGPT players arrive already sold
The players ChatGPT sends behave differently from everyone else’s.
| Source | Clicked Play | Played 7+ days |
|---|---|---|
| ChatGPT | 47% | 11.5% |
| 44% | 8.2% | |
| 42% | 5.1% |
Clicked Play is site visitors since 1 August. Played 7+ days is players who joined since June and came back on seven or more separate days.
They click through slightly more than Google’s visitors, and they stay more than twice as well as Reddit’s. My guess is that ChatGPT knows the game by now, from what it’s read about it, and describes it well enough that the people it sends already know what they’re getting. I can’t prove that. My analytics only sees browsers that run JavaScript, so OpenAI’s crawlers never show up in it, and I can’t tell a search from a memory. It’s 96 players so far, which isn’t many, but of the three big sources it’s the one whose players stay.
Why I’m not pushing for more players
Chris gave me an analogy I keep coming back to: it’s a leaky bucket. There’s no point pouring more people in the top while they’re falling out of the bottom. The June push, when I launched the German version on Reddit, brought 168 new players in two weeks, seven times the normal rate. Ten weeks later, six of them are still playing.
So I’m not chasing a thousand daily players. The game still has holes, and around nine new players a day arriving on their own is enough to test every change I make to the first session. Most of the organic traffic took about three months to arrive, and it started after the SEO work I was dubious about.
The difference now is that it keeps going without me. Before, every new player came from me pushing something. Now the game has its own wheels: a steady flow of players arrives whether I post or not. I’ve lost plenty of players along the way, and that’s fine. When I do the next marketing push, it lands on top of that flow instead of starting from nothing, so I expect it to move things a lot faster than June’s did. And there’s still plenty of SEO left to do: tightening the wording on the pages, listing on more directories, and getting posts like this one onto other sites.
What I’d tell anyone building a game
- Connect Google Search Console the day you launch. I waited until 30 June, so I have no record of how Google indexed the site before that.
- Get the first players by hand. Message people, post where your kind of player already hangs out, and watch what they do. You’re after learning, not numbers.
- Build pages for what people already search for. Compare your game with the ones your players already know. It felt old-fashioned and took about three months, but that’s when Google and ChatGPT started sending players.
- Ask every new player where they found you. Tracking misses a lot, especially ChatGPT, where people read the answer and then search for your name.
- List the game on every directory that fits. itch.io, IGDB, IndieDB and browsergamestop all list mine, and new game indexing sites appear every week. Each one is another link back to the game.
- Set up a press kit before you submit anywhere. One page with everything a site might ask for: screenshots, logos, a short and a long description, a video, the key facts. When a directory wants something, you copy it from there instead of digging through folders on your computer. Here’s ours.
- Fix the leaky bucket before you pour more in. A push gets you a spike. If the first session doesn’t hold people, most of them are gone within weeks.
- Have someone outside the project to ask. Chris is a friend who works in marketing, and he often points out the obvious thing I’ve been too close to see.