← ContentsEssay II

how Quicko’s tools pages earned 71% of our search clicks

SEO, citations, and what AI was actually good for.

4 min readFiled under: SEO, AI, tax season

SEO has a funny way of testing your patience. You can do everything right, tick every box, add every optimisation on your pages, and still wait months to see meaningful organic traffic. So, you start looking for ways to bring people to your website before they’re ready to buy anything.

At Quicko, we had a particular reason to care about this. We’re a tax-filing platform in India, which means a large chunk of our traffic arrives in the three months around the ITR filing deadline. The rest of the year is quieter. We invested in top-of-funnel (TOFU) content: blogs, guides, community forums and more than 30 free tools on tools.quicko.com so people would meet Quicko well before they were ready to file.

The problem was that tax rules change, APIs update, and tools break. Some of ours had been discontinued, with a banner informing visitors that they were no longer operational. Rebuilding them kept getting delayed; production bugs, product updates and new features had a way of winning the priority argument.

We’d started rebuilding the tools in December 2025, using tools like v0 to cut development time from weeks to days. But with more than 30 tools to bring back, progress was gradual. By May, just two months before the ITR filing deadline, we found that the old subdomain was still generating over 300,000 impressions. Google hadn’t forgotten those pages. We just hadn’t given their visitors a reason to stay.

Why moving the tools wasn’t enough

The new tools found a home on quicko.com/tools, but the old URLs on tools.quicko.com continued to rank. Two sets of pages were competing for the same searches, and only one was where we wanted people to land. The priority was to move the remaining tools to the main domain before filing season peaked.

We no-indexed the pages on tools.quicko.com and embedded the tools into the pages on the main domain, where all the supporting content lived. This gave visitors the tool and the information they needed in one place.

But moving a URL is the easy part. Making the page worth visiting was the challenge.

Balancing content with the right amount of SEO

As I migrated the tools, I started reading the pages the way a stranger would: what does this answer, what does it miss, and does the structure make sense to someone who hasn’t spent six months thinking about tax tools.

I started with the PAN–Aadhaar link status tool, one of our busiest pages. The page already had enough content: linking statuses, inoperative PANs, deadlines and exemptions. But the way it was structured was doing very little work.

“What is it?” could mean almost anything. “What is PAN–Aadhaar linking?” doesn’t leave much to anyone’s imagination. I rewrote the headings to name the subject explicitly and reflect how people searched for it. No keyword stuffing, no SEO incantations. Just clearly stating what each section was about.

The FAQs got the same treatment. Rather than repeating the main content, they covered specific situations: a PAN that still showed as inoperative after linking, or questions an NRI might have. The main headings covered the big questions; the FAQs picked up the niche, little ones. The aim was to make the existing information easier to find for users and search engines to understand and use, without turning the page into an encyclopaedia.

If you’re citing the law, cite the law

The tricky thing about tax content is that rules change and references get outdated. An outdated citation can sit inside a perfectly polished paragraph and look entirely convincing. So source-checking had to be part of the process, with legal claims pointing to the relevant legislation or official notification.

As search evolves beyond blue links into AI-generated answers, traceability matters, too. Citations help readers verify claims and give LLMs a path back to the underlying evidence when retrieving information. They don’t guarantee AI visibility, but publishing claims without sources gives both readers and the systems using our content less to work with.

If we’re going to tell people what the law says, we should show our working.

Twenty-five pages shouldn’t mean twenty-five separate reinventions

By the third or fourth page, I was having the same conversation with myself. Over and over again. H1s, FAQs, worked examples, citations. Surely I could outsource the repetitive parts of it to a machine. So I turned the page structure into a Claude skill.

It pulls query data from Google Search Console, related terms from Keyword Planner, and competing pages. From that it finds content gaps, maps queries to headings, and drafts FAQs from what people actually search, all within the structure I’d defined. Instead of starting with a blank document, I have a first draft grounded in data.

The skill does the legwork. I decide what’s relevant, what belongs on the page, and whether the draft is actually any good. I review the output, check the sources and have the hard facts verified. We’re automating the grunt work, not outsourcing our brains.

So, did it work?

Between May and September 2026, Quicko received 955K clicks from Google Search. The tools pages contributed 674K of them, 71% of all clicks, from just 29% of impressions. Their CTR was around 6%, against 1% for the rest of the site.

The PAN–Aadhaar page alone brought in 508K clicks, compared with 78K for the homepage. These numbers show the scale of the tools’ organic traffic, not the impact of my changes alone. The pages moved at different times, and they benefited from a government deadline that drove demand. What we could control was what happened when people searched and landed on our page: whether the tool worked, whether the content answered their question, and whether the information held up to scrutiny.

We also reworked more than 25 pages in under a month, with a process we could reuse instead of rebuilding the brief from scratch every time.

For me, that’s the usefulness of AI in content. Not getting a machine to churn out 25 pages, but giving it a process worth repeating.

AI can generate the words. Deciding what deserves to be said is still the job.