Finds problems on your site, explains them in plain words, shows the proof — and re-checks after you fix: fixed, not fixed, or couldn't confirm. No score out of 100. 70+ checks, nothing to install, free right now — no credit card, no trial that runs out.
What could you build with GPT-5.6 that wasn’t practical before?
Maker
Honestly? A one-person product with a QA discipline I couldn't have afforded before.
Koval SEO Console is an SEO checkup that explains problems in plain words, with proof from your own pages, and re-checks after you fix. GPT-5.6 Codex agents wrote it end to end — crawler, rules engine, UI.
But the part that wasn't practical before is the proving, not the code: every check had to catch a real bug we planted on a live test site before it shipped. Spec research, the rule itself, evidence capture, live verification — per check, that used to be a team's week. With GPT-5.6 it became a loop one person can run: 70+ proven checks in a few weeks, and the 200+ on the roadmap is realistic for the same reason.
And agents don't get tired of verification. The same loop that built the product now guards its promise — nothing ships unless it demonstrably catches what it claims to catch.
Report
Maker
📌
Hi Product Hunt! I'm Oleh, and this is Koval SEO Console.
I didn't set out to build an SEO tool. I had my own site, felt like Google was ignoring it, and ran it through the usual checkers. What I got back was a 73 out of 100 and a wall of warnings I couldn't read. I remember sitting there thinking: I don't want a grade, I want someone to tell me in normal words what's wrong with my own site — and later, whether my fix actually worked. Nobody would. That frustration is the product.
So Koval SEO checks your site and explains every problem the way you'd explain it to a friend: what's wrong, why it matters, and how sure we are — with the proof from your own pages sitting right next to it. Priorities are written in words too. No score out of 100 anywhere.
The approach changed a lot while building. I started out thinking the hard part was finding problems. It isn't — every tool finds problems. The hard part is being believable about them. That realization reshaped everything: the full list of checks became public, including the ones we deliberately don't run because there's no reliable way to run them. And re-checking got the strictest rule in the product: if a problem simply disappears, Koval SEO won't call it fixed — you get "couldn't confirm," because gone isn't the same as fixed. Building a tool that's allowed to say "we're not sure" turned out to be the whole point.
Koval SEO is free right now — the whole product, no credit card. If you try it on your site, tell me which explanation reads confusing — a confusing sentence is exactly the kind of bug I most want to fix. I'll be here all day.
Report
Congrats on the launch, Oleh. Nice to see a solo maker with a clear opinion about how a tool should talk.
I ran the preview on planmoon.app. It scanned 24 pages out of ~258, 4 checks passed, 59 could not be checked.
The text below explains this very well. It says it is not a site-health verdict, and it says why AI-crawler access cannot be tested from a browser. All fair.
But the first thing I read was the title: "Good news — nothing critical." And I felt relieved for a second. Then I read the rest and understood that almost nothing was actually checked.
Your strictest rule is "gone is not the same as fixed." I think the preview title is the one place where the tool sounds more sure than the evidence. Maybe something like "4 checks passed, 59 not evaluated yet" says the same thing without the relief.
Will make a free account and run the full audit later. Good work.
Report
Maker
@saied_alimoradi This is exactly the feedback I was hoping for — thank you. You're right, and you caught the product breaking its own rule: "nothing critical" is a bigger claim than 4-of-63 checks can back. The title should carry the scope, not just the mood — changing it to lead with the numbers, something like "4 checks passed — 59 need the full check," so the relief can't arrive before the context. Shipping that fix today.
And yes — the free account runs the full thing: all checks, across the whole site, with the re-check loop. I'd genuinely like to hear if it holds up to the same standard.
Congrats on the launch, Oleh. Nice to see a solo maker with a clear opinion about how a tool should talk.
I ran the preview on planmoon.app. It scanned 24 pages out of ~258, 4 checks passed, 59 could not be checked.
The text below explains this very well. It says it is not a site-health verdict, and it says why AI-crawler access cannot be tested from a browser. All fair.
But the first thing I read was the title: "Good news — nothing critical." And I felt relieved for a second. Then I read the rest and understood that almost nothing was actually checked.
Your strictest rule is "gone is not the same as fixed." I think the preview title is the one place where the tool sounds more sure than the evidence. Maybe something like "4 checks passed, 59 not evaluated yet" says the same thing without the relief.
Will make a free account and run the full audit later. Good work.
@saied_alimoradi This is exactly the feedback I was hoping for — thank you. You're right, and you caught the product breaking its own rule: "nothing critical" is a bigger claim than 4-of-63 checks can back. The title should carry the scope, not just the mood — changing it to lead with the numbers, something like "4 checks passed — 59 need the full check," so the relief can't arrive before the context. Shipping that fix today.
And yes — the free account runs the full thing: all checks, across the whole site, with the re-check loop. I'd genuinely like to hear if it holds up to the same standard.