Unfiltered essays, teardown playbooks and case studies from a small studio in Delhi and Mumbai. No thought-leadership. No listicles. Only what we would tell a friend over coffee.
We ran the same brand through six AI website builders and our own studio process. Here is the honest scorecard on SEO, conversion, Core Web Vitals and the year-two cost nobody talks about.
After auditing 42 homegrown D2C storefronts in 2026, the pattern is embarrassingly consistent. Here is what we change first.
Our brand playbook for professional-services firms who have been told they need to "look more modern".
Our internal timeline, the meetings we refuse to take, and the one Figma file rule that saves a full week.
How a private aviation client used one focused microsite — not a full site rebuild — to change the conversation with HNW clients.
The one-hour AI site looks like a gift. It becomes a bill when you try to change anything.
We ran the same brand through six AI website builders and our own studio process. Here is the honest scorecard on SEO, conversion, Core Web Vitals and the year-two cost nobody talks about.
Every founder we met in Q2 2026 asked the same question in the first five minutes: "Why should I hire a web design agency in India when an AI website builder gives me a live site in eleven minutes?" It is a fair question and the honest answer is not the one most agencies give. So we ran the experiment. We took a real client — a Bengaluru D2C skincare brand with 4,200 monthly organic visitors — and rebuilt their homepage six times: five leading AI website builders (Framer AI, Wix ADI, Durable, Hostinger AI, and one we will not name because their legal team is enthusiastic) and once in our own studio. Same brief. Same copy deck. Same product photography. Then we shipped all six to staging, ran Lighthouse, PageSpeed, Ahrefs, and a 400-person unmoderated user test.
The AI-generated sites averaged a Lighthouse performance score of 61 on mobile. Ours scored 96. That gap is not because AI builders write bad code — most of them ship reasonable HTML. It is because they ship every possible component on every page, load four web fonts, embed a video hero that autoplays on 4G, and inject a chat widget the client did not ask for. An agency site is a subtractive act. An AI site is an additive one.
SEO was the more interesting split. All six sites were indexed within 72 hours. But when we tracked keyword coverage after 60 days, the AI sites ranked for an average of 14 branded queries — the brand name, the founder name, small variants. Ours ranked for 217 queries, 80% of which were non-branded commercial-intent terms like "vitamin C serum for oily skin India" and "clean beauty brands Bengaluru". The difference was not magic. The AI builders generated generic H1s ("Welcome to [Brand]"), duplicate meta descriptions across pages, and zero internal linking strategy. We wrote 43 pages of content architecture before a single pixel moved.
Conversion was closer than we expected and worth being honest about. On the homepage-to-PDP click, the AI sites averaged 22%. Ours hit 31%. Meaningful, but not the 3x gap agency Twitter would have you believe. Where the gap widened brutally was PDP-to-checkout: AI sites averaged 4.1%, ours hit 9.7%. The reason is boring — the AI builders defaulted to a Shopify template checkout with six form fields above the fold. We rebuilt it as a two-step, Indian-payments-first flow with UPI Intent as the primary CTA. That single decision moved more revenue than the entire visual redesign.
Now the part nobody writes about: year two. We audited eleven sites this year that were originally generated by AI builders in 2024–2025. Every one arrived with the same brief — "just small tweaks" — and every one required a full rebuild. AI-generated sites are a snapshot, not a system. There are no design tokens, no component library, no naming convention, no reusable patterns. Change a button colour on the third card of the second section and you are editing forty places by hand. The one-hour site becomes a forty-hour bill the first time your brand evolves.
None of this is an argument against AI. We use AI every single day — for copy drafts, for alt text, for image cleanup, for code review, for the boring 30% of production. It is an argument against confusing generation with design. Generation is a moment. Design is a system that has to survive two years of small edits by a rotating cast of humans who do not remember why the padding is 28px.
So here is the honest recommendation, and it is not the one that fills our pipeline. If you are pre-revenue, testing a positioning, or shipping a landing page for a Meta campaign that dies in six weeks — use an AI website builder. Framer AI in particular is genuinely excellent for this. If you are a D2C brand doing more than ₹50 lakh a month, a professional-services firm charging more than ₹5 lakh per engagement, or a founder-led startup where the website is the first sales conversation — hire a studio. Not because we are precious about craft. Because the compounding cost of a bad system, over two years, is the salary of a full-time designer you never hired.
The trending question is really the wrong question. It is not AI website builder versus web design agency in India. It is: how long does this website need to survive, and who edits it in month fourteen when the founder is on a flight and the intern needs to add a new product category before Diwali. Answer that first. The tool falls out of the answer.
“An AI site ranks for your name. An agency site ranks for your market.”
After auditing 42 homegrown D2C storefronts in 2026, the pattern is embarrassingly consistent. Here is what we change first.
We spent a rainy weekend in Bandra pulling apart 42 Indian D2C sites — skincare, coffee, apparel, small-batch spirits. Nine of them had a founder story worth reading. Two had a checkout that did not make us wince. None had a product page we would frame.
The problem is almost never taste. Indian founders have great taste. The problem is that the site was built by whoever was cheapest on a Shopify marketplace, and the theme was tuned for a Los Angeles athleisure brand from 2021.
The six fixes we run on day one: kill the auto-rotating hero, replace stock lifestyle photography with product-in-hand shots, tighten body copy to 62 characters per line, rebuild the PDP above the fold, remove every gradient button, and rewrite the shipping strip in the founder's own voice. That is usually a two-week sprint and it moves conversion 18–34% before we touch anything structural.
None of this is clever. It is just the discipline of noticing.
“Premium is not a font. It is the silence between elements.”
Our brand playbook for professional-services firms who have been told they need to "look more modern".
Every quarter a CA firm founder walks into our Delhi studio and says the same sentence: "We want to look more modern." Nine times out of ten what they actually want is to stop being embarrassed when a 32-year-old founder client Googles them.
Modern is the wrong word. The word is composed. A composed brand for a professional-services firm is not a wordmark refresh — it is a room the client walks into, digitally. Typography that suggests the partner reads. Photography that shows the office actually exists. A services page that names the two things the firm is genuinely great at, and quietly omits the rest.
We shipped this for a 34-partner firm in Nariman Point last quarter. Their inbound from founder-led startups tripled in 90 days. The logo did not change.
Our internal timeline, the meetings we refuse to take, and the one Figma file rule that saves a full week.
Twenty-one days sounds aggressive until you see the meetings we do not take. No kickoff call over 45 minutes. No weekly status meeting. No design review that includes the client's second cousin. We replaced all of it with a shared Loom and a single decision-log doc.
Week one is discovery and copy — yes, copy first. Week two is design in a single Figma file that nobody outside the pod can edit. Week three is build, QA, and launch. That is it.
The single Figma file rule is the one that surprises people. Multiple files means multiple sources of truth, means the developer builds from a stale frame, means a week of rework. One file. Named sections. Every screen visible on one canvas.
“The single Figma file rule saves us a week on every project. No exceptions.”
How a private aviation client used one focused microsite — not a full site rebuild — to change the conversation with HNW clients.
The client operates a small fleet out of Delhi and Mumbai. Their existing site was fine — corporate, safe, forgettable. Their sales team was losing deals in the first meeting because the site did not match the price point the client was quoting.
We shipped a 4-page microsite in 19 days. Not a rebuild. A parallel property, indexed separately, that the sales team sent as a link before every discovery call. Full-bleed cinematography, a single-column scroll, no navigation until the last section, and one CTA: "Request a private walkthrough."
In the six months after launch, the microsite was credited on ₹18 Cr of closed charter revenue. The main site is still corporate and safe. It does a different job.
The one-hour AI site looks like a gift. It becomes a bill when you try to change anything.
We have now inherited 11 sites in the last year that were originally generated in an hour by an AI builder. Every single one arrived with the same brief: "We just need small tweaks." Every single one required a full rebuild.
The reason is boring. AI-generated sites are a snapshot, not a system. There are no components, no tokens, no naming convention, no reusable patterns. The moment you want to change a colour on the third card of the second section, you are editing 40 places by hand.
This is not an argument against AI. We use AI every day. It is an argument for treating your website like a system that has to survive two years of small edits by a rotating cast of humans. That is a design problem, not a generation problem.