An Indian AI course company pushed its site into the global top 100k with "Diwali photo prompts"

Build Fast with AI, a Bengaluru company, sells $399 live AI courses. Its site climbed from the global top 50 million to the top 84k in 90 days, and what drove it was not the courses but a prompt collection for ordinary Indians editing festival photos. The playbook is easy to understand and easy to copy; the hard part is what comes after: turning the traffic into money. Below: what it did, where the traffic came from, where the money comes from, and whether you should follow it.

A post titled "written for search"

On Sep 11, 2026, blog.buildfastwithai.com published "100 Trending AI Photo Prompts 2026 (updated weekly)". The intro covered the "1980s retro" photo trend popular in India at the time: users turned their current selfies into 80s family portraits, Bollywood-era studio shots, and Kerala-style old photos. The third paragraph was blunt: "this collection is written around search intent," and then it listed AI photo prompts, Gemini image prompts, and Nano Banana 2 prompts in one breath.

Over the next two months, the series ran almost weekly: 9-14 covered Gemini photo prompts, 9-17 covered "childhood vs now," 9-24 covered saree and traditional outfits, 9-26 covered wedding couple photos, 10-14 covered Navratri, 10-18 covered Dussehra, and 11-02 covered Diwali (dates from the blog sitemap). The Diwali post opened by giving the 2026 Diwali date as Nov 8, then offered 50 copy-paste prompts, such as "using my uploaded photo, dress me in traditional [color] clothing, with lit brass oil lamps and marigolds beside me... the face must stay exactly the same, 4:5."

Over the same period, the domain's global rank moved like this: 9-10 rank 340k, 10-01 rank 173k, 11-05 rank 84k. In Chrome real-user data, the India region was in the top 50k tier in August and jumped into the top 5k tier by September; the US stayed in the top 1M tier, and other countries barely moved. Almost all the traffic came from India.

What it actually is

The website and blog of an AI education company. Founder Satvik Paramkusham graduated from IIT Delhi and previously did developer relations at Portkey. The homepage reads "50,000+ learners, 400+ workshops, 100+ companies," with the logos of Google for Developers, BCG, and Accel. They sell three things:

• A two-week live crash course for FDEs (Forward Deployed Engineers), $399 (struck-through original price $500);
• three-week Claude Mastery live course; early-bird $100,11-08 off until the deadline;
• Corporate training and AI consulting.

Plenty of companies sell courses, but this one stands out for its blog. Every page in the blog's nav bar links to the FDE course, with the tracking tag "blog_navbar"; article footers carry workshop signups and a "70,000+ community" link. It also runs an AI tool directory called BuildFast Studio, plus small toys like a résumé roaster and daily quizzes. These exist to drive traffic, not as the main traffic source.

Where the traffic comes from: three lines running together

Line one: mass-market photo prompts. This is the series from the intro. Using Gemini and ChatGPT to edit photos is a nationwide hobby in India, and Navratri, Dussehra, and Diwali all fall between 9 and November. Searches for these terms far outnumber searches for "Kimi K3.1 review." India traffic jumped in September, right when this series began. Outsiders can't see the ranking for each keyword, so this link is an inference.

Line two: riding new model launches. On 11-06 it published two reviews, one on ChatGPT Images 2.6 and one on Gemini Omni 1.2 Flash. By sitemap count, the blog published only a few to twenty-odd posts per month in 2025; starting March 2026 it rose to 40 to 70 per month, and by September it hit 117 (this figure includes updates to old posts, so it's approximate). But searching "kimi k3.1 pro review" or "best ai models october 2026" in the India region, it doesn't appear in the top ten. Analytics Vidhya, MindStudio, and others compete for these terms, and publishing volume alone doesn't get you ranked.

Line three: programmatic long tail. The main site has about 550 "prompt library for [profession]" pages, about 340 AI tool category pages, and list pages such as "AI courses for doctors / accountants / teachers / kids," "free Indian government AI courses," and "AI courses in Hindi," totaling about 1,500 template pages. The sitemap was still growing as of 11-06.

There are also some icing-on-the-cake moves: a YouTube channel with 108k subscribers; and in September, a hackathon on Unstop, the competition platform Indian students use. Registration closed 9-25, with a prize of ₹18,000 plus up to 10 internship spots, and registrants received a shareable "Builder Pass" poster. Registration numbers weren't disclosed, so the contribution is unclear. No paid ads turned up in searches, but that doesn't mean none were run.

Where the money comes from: the people reading and the people buying are different groups

The only visible revenue is courses and corporate training. The problem: someone searching "Diwali photo prompt" is an ordinary person who wants to add a diya to a family photo, while someone buying the $399 FDE course is an engineer looking to change careers. The two groups barely overlap. So a 6x rise in rank doesn't mean course sales grew 6x. The company has never disclosed revenue, and the founder hasn't shared numbers. The likely role of the photo-prompt traffic is to build authority for the whole domain, making the comparison posts and course pages easier to find in search (inference).

Who's fighting for the same ground

The most telling is promptpin.in: a new site focused only on festival prompts that climbed from outside the rankings to rank 90k within 117 days. That shows newcomers can get into these terms, and also that many have already gotten in.

What can be copied

The product itself has nothing worth copying: courses, prompt collections, and AI news are crowded in India. What's worth copying is the distribution combination:

1. One domain hosts two kinds of content: new-model reviews for developers (smaller audience, higher willingness to pay) and festival and trend photo prompts for the general public (large volume, strongly seasonal).
2. Publishing schedules follow the festival calendar, with the year in the titles. Actually, they published late: the Navratri post went up 10-14, after the festival had already started; the Dussehra post (10-20) went up 10-18, and the Diwali post (11-08) went up 11-02, only a few days ahead. Publishing two to three weeks early gives search engines time to index the page, which is exactly where a follower could outdo them (inference).
3. Every post has a monetization hook with tracking tags, so it's clear which post brought in course buyers.
4. Templates like "profession × prompt" and "audience × course" generate hundreds or thousands of long-tail pages, catching scattered searches for the whole domain.

If it were you

C: Worth studying the playbook, but not worth copying on its own. The traffic is real, but it rides an Indian mass-search season. These terms are already crowded by big media outlets, pure prompt sites, and Google itself. Worse, the company has never shown that this traffic turns into money.

• Who it suits: small teams that already have something to sell (a course, SaaS, templates, or a tool) but lack free traffic, especially those working in languages like Indonesian, Vietnamese, Portuguese, or Thai, where lots of people use AI to edit photos but local-language prompt content is still thin (inference; I haven't checked competition in these languages).
• Who it doesn't suit: people with nothing to sell who want to earn from ads on a prompt site (ad rates for India and Southeast Asia traffic are low, so this is likely a site with impressive traffic and thin revenue, an estimate); and anyone looking to compete head-on in English or the Indian market against Analytics Vidhya or Times of India.
• How to do it: pick one language market and map out a year of festivals and trends (Vietnamese Tet 2027-02-06, Indonesian Eid al-Fitr 2027 in March 2027, Brazilian Carnival in early February 3). Publish a "4 copy-paste photo prompts" post 50 weeks before each festival, with the year and model names (Gemini, ChatGPT, Nano Banana) in the title. Between festivals, write reviews whenever new models launch. Then use templates to spread out a long tail of "[profession / audience] × AI" pages. Put an entry point to your own product on every page.
• How to get users: rely mainly on organic search. Pair each post with a short demo video (before / after) posted to YouTube Shorts and Instagram Reels, linking back to the article. To give it a push, run a small local-student hackathon on the platform students use there, and create a shareable poster.
• When to start: to catch the 2027 Lunar New Year and Eid season, get the site built by late December 2026, and publish the first batch of festival posts by mid-January, giving search engines time to index them (inference).
• Signs you've misjudged it: (1) after Diwali (11-08) passes, its global rank drops back beyond 200k in December, showing this wave was a one-off festival spike; (2) conversely, if rank holds within 100k after the festival and it starts publicly disclosing course sales, the playbook is worth more than it looks and the rating should go up; (3) if Google makes "festival photos" a one-click template in Gemini and rolls it out across countries, search demand for mass-market prompt articles will be eaten away.