Indian startups vs US SaaS: how ready are their websites for AI search?
110 Indian startups
3%
open their homepage with a sentence that says what the company does
"Mann hai toh money hai"
110 US SaaS companies
21%
open their homepage with a sentence that says what the company does
"Dub is the modern link attribution platform for short links, conversion tracking, and affiliate programs."
We ran our SaaS homepage audit unchanged on 110 Indian startups, from listed companies to seed-stage AI tools. On the technical checklist the two samples are close: 75% of Indian homepages carry Organization schema, against 74% of US SaaS. The gap is in the opening copy, and in how many homepages a labelled bot can load at all: 17 of 110 turned us away in India, 6 of 110 in the US.
The samples differ. The US set is all B2B software. The Indian set spans fintech, consumer apps, D2C brands, edtech, health, SaaS and AI. So wherever it applies, we also mark India's 35 SaaS and AI companies on their own, the closest like-for-like comparison.
Where the two samples split
Every check, sorted by the size of the gap. Share of homepages that loaded: 93 in India, 104 in the US.
| Check | All Indian startups | Indian SaaS and AI only | US SaaS |
|---|---|---|---|
| Meta description | 98% (91 of 93) | 97% (31 of 32) | 99% (103 of 104) |
| Exactly one H1 | 73% (68 of 93) | 81% (26 of 32) | 79% (82 of 104) |
| Any JSON-LD | 80% (74 of 93) | 78% (25 of 32) | 78% (81 of 104) |
| Organization schema | 75% (70 of 93) | 78% (25 of 32) | 74% (77 of 104) |
| Publishes llms.txt | 50% (53 of 105) | 56% (19 of 34) | 69% (75 of 109) |
| Opening says what the company does | 3% (3 of 93) | 9% (3 of 32) | 21% (22 of 104) |
| "Company is a..." definition | 1% (1 of 93) | 3% (1 of 32) | 9% (9 of 104) |
| Wikidata link in sameAs | 1% (1 of 93) | 3% (1 of 32) | 13% (13 of 104) |
| FAQPage schema | 10% (9 of 93) | 13% (4 of 32) | 8% (8 of 104) |
Only Fractal, Exotel and ToolJet open by naming themselves and saying what they do. We read every extracted opening by hand; counting the borderline ones too (a "Brand:" headline, a sister brand's name) lifts India to 8% at most. 12 Indian homepages sent no H1 at all in their HTML, against 2 in the US, so their opening exists only after JavaScript runs. Many of the rest lead with an offer or a slogan, while the meta description underneath already says what the company is.
17 of 110 doors stayed shut
What happened to each homepage request from a polite bot that names itself.
Indian startups17 of 110 not loaded
Indian startups: 93 loaded, 13 refused by bot protection, 4 never answered.
US SaaS6 of 110 not loaded
US SaaS: 104 loaded, 6 refused by bot protection, 0 never answered.
Refused (13): Swiggy, Naukri, Pine Labs, Meesho, Capillary, Wakefit, Thyrocare, LeadSquared, Scapia, Leap Scholar and HealthifyMe answered with a 403. RateGain and Zepto served a bot-protection page in place of the homepage (an F5 rejection and an AWS WAF challenge), which we count the same way.
Never answered (4): Zomato, Nykaa, Policybazaar and MobiKwik. Two timed out, one sent response headers too large for a standard client, and one has a certificate chain that strict clients reject.
Robots.txt is not the problem. Of 96 readable Indian policies, only IndiaMART, Snitch and Kuku FM block any of the four big AI crawlers. If your firewall is stricter than your robots.txt, the crawler you invited is still turned away.
| Crawler | Named, India | Blocked, India | Named, US | Blocked, US |
|---|---|---|---|---|
| GPTBot | 16% | 2 | 21% | 2 |
| ClaudeBot | 15% | 3 | 21% | 2 |
| PerplexityBot | 16% | 2 | 21% | 0 |
| Google-Extended | 17% | 1 | 18% | 1 |
llms.txt, and who wrote it
7 of India's 53 files are Shopify's default "Agent Instructions" file, which explains how to shop the store but never says what the brand is: Mamaearth, boAt, Sugar Cosmetics, Bombay Shaving Company, Snitch, Bluorng and Bombay Sweet Shop. Without them, India is at 44%. And 40% of Indian files open with "Company is a...", against 75% in the US.
Indian startupsout of 105
US SaaSout of 109
India: 46 company-written llms.txt files, 7 Shopify defaults, 52 with none. US SaaS: 75 with an llms.txt, 34 without.
By stage, side by side
Listed Indian companies are the hardest to read: 12 of 30 refused us or never answered. Early-stage Indian startups are the only group level with their US peers.
| Stage | Loaded | llms.txt | Organization schema | Direct answer | Median HTML |
|---|---|---|---|---|---|
| Listed | 18 of 30 | 31% vs 66% | 83% vs 83% | 6% vs 14% | 350 vs 610 KB |
| Late-stage private | 30 of 32 | 48% vs 69% | 77% vs 76% | 0% vs 18% | 405 vs 757 KB |
| Growth | 26 of 29 | 66% vs 83% | 65% vs 75% | 4% vs 38% | 365 vs 392 KB |
| Early | 19 of 19 | 58% vs 56% | 79% vs 56% | 5% vs 17% | 329 vs 267 KB |
By category, against the US SaaS average
The US sample has no categories to match, so each Indian category is set against all US SaaS. Groups have 11 to 22 companies, so read them as directional.
| Category | Loaded | llms.txt | Organization schema | Direct answer | Median HTML |
|---|---|---|---|---|---|
| Fintech | 17 of 21 | 58% vs 69% | 88% vs 74% | 0% vs 21% | 338 vs 536 KB |
| SaaS | 19 of 22 | 48% vs 69% | 74% vs 74% | 11% vs 21% | 361 vs 536 KB |
| Consumer internet | 12 of 18 | 31% vs 69% | 75% vs 74% | 0% vs 21% | 166 vs 536 KB |
| D2C brands | 12 of 13 | 85% vs 69% | 83% vs 74% | 0% vs 21% | 989 vs 536 KB |
| Edtech | 10 of 11 | 36% vs 69% | 60% vs 74% | 0% vs 21% | 480 vs 536 KB |
| Health | 10 of 12 | 25% vs 69% | 50% vs 74% | 0% vs 21% | 333 vs 536 KB |
| AI-native | 13 of 13 | 69% vs 69% | 85% vs 74% | 8% vs 21% | 329 vs 536 KB |
D2C brands look strong on llms.txt only because of Shopify's default file, and their homepages are the heaviest of any group. Fintech has the most complete schema and not one opening that says what the company does.
English first, rupees often, Indian languages rarely
39% of the Indian homepages that loaded show a rupee amount, from loan limits to discount codes.
9of 93 have text in an Indian script or an alternate page in an Indian language: IndiaMART, Go Digit, Shiprocket, Pristyn Care, Jar, Pocket FM, Kuku FM, Smallest.ai and SpeakX.
Hindi written in Latin script, like Moneyview's headline, is not counted. 7 homepages declare
Indian English (en-IN) and 17 declare US English, mostly SaaS and AI companies
selling abroad; the rest say plain en or nothing. For companies whose users read Hindi, Tamil or
Bengali, there is very little native-language text on the homepage for an answer engine to find.
What Indian founders can fix this week
- Put your meta description on the page. If it already says what you do, move that sentence into the first paragraph, with your company name at the start, above the offer or the slogan.
- Ship your opening in the HTML. If your H1 and first paragraph only appear after JavaScript runs, most AI crawlers see an empty page. Server-render the top of the homepage.
- Let the bots you allow through the firewall. Robots.txt rules mean little if Cloudflare, Akamai or AWS WAF turns crawlers away first. Check your bot settings against the AI crawlers you actually want.
- Replace the platform llms.txt. If you're on Shopify, your llms.txt may be the default agent file. Add a line at the top that says who you are and what you sell.
- Get a Wikidata entry and link it. Only 1 of 93 Indian homepages link one. For a listed company or a funded startup with press coverage, it's an afternoon's work.
For how answer engines pick what to quote, see GEO: showing up inside AI answers. If you're weighing an Indian studio against a US agency for the work, we've compared the two.
How we did it
We picked 110 Indian startups and scaleups by hand to spread across stage (listed on NSE or BSE, late-stage private, growth, early) and category (fintech, SaaS, consumer internet, D2C brands, edtech, health, AI-native), drawing names from Indian startup IPO trackers, unicorn lists and YC-backed Indian companies. Companies already in the US SaaS sample (Freshworks, Postman, Plane) were left out. It is a deliberate spread, not a random sample.
We then ran the SaaS homepage audit unchanged: three
requests per company, one at a time, for the homepage, /robots.txt and /llms.txt,
with no logins, retries or JavaScript, and a user agent that names us. Fetched from India on 17 and 18
October 2026, so sites served their Indian versions.
"Opening says what the company does" is the same heuristic as the SaaS audit: the H1 or first real paragraph must define the company ("Company is a...") or open a sentence with its name as the subject of a verb. Stage labels are our own classification from public listing and funding information.
