Skip to content
Mutual funds

How Indian investors research funds before buying

An assistant can restate your expense ratio accurately and your three-year return badly. That difference decides most of what it says about your schemes.

5 min read · Updated 2026-08-27

Indian investors do most of their fund research before they ever open an app, and a growing share of it now happens inside an AI answer rather than a search results page. An assistant can restate your expense ratio, lock-in and tax treatment accurately, because those are defined facts with a regulator behind them. It cannot reliably state your returns, because a return is only true for a window, a benchmark and a date, and the engine is quoting whatever it last read. Publish the durable half as text, and the fragile half with its basis attached, and you control most of what an assistant says about your schemes.

The decision is made at the research stage

The Association of Mutual Funds in India put industry assets at ₹85,75,657 crore as on 31 July 2026, and the same month's SIP collection at ₹31,961 crore. Those numbers describe money already committed. The more useful figure for a marketing lead sits in the Securities and Exchange Board of India's Investor Survey 2025, which found 53 percent of Indian households aware of mutual funds or ETFs and 6.7 percent actually holding them.

The gap between those two is your market: roughly eight households who recognise the category for every one invested in it. Each of them is somewhere in a research process that ends in an app and starts a long way before it. What has changed is where the middle of that process now happens.

Google's own measurement, published on its blog in May 2025, is that AI Overviews drive over a 10 percent increase in usage of Google for the query types that show them, in its biggest markets, naming the United States and India. The research stage did not move to a new website. It moved into a generated paragraph.

What does an investor actually ask before they buy?

One line typed into an assistant is not one question. A prompt such as best mutual funds to invest in India for long term fans out into six: returns over 3, 5 and 10 years, expense ratio, the fund manager's track record, exit load and lock-in, risk grade and volatility, and tax treatment. The engine builds its answer from whichever of the six it can verify, and quietly drops the ones it cannot.

SEBI's Investor Survey 2025 asked investors what information they look for. Among mutual fund and ETF investors, 75 percent named performance, 52 percent named cost, and 11 percent named the fund manager's track record.

That ordering inverts most fund-house marketing. Cost is named by roughly five times as many investors as the manager story, and the manager story is usually where the brand budget goes. An assistant answering a fund question behaves more like the survey than like your brochure: it leads with performance, reaches for cost second, and mentions the manager only if a source hands it one.

The half an assistant answers well

Some scheme facts are stable, defined and backed by a regulator. Expense ratio. Exit load. Lock-in period. Minimum investment amount. The riskometer band. Tax treatment for the scheme type. None of these are opinions, and none of them move between one crawl and the next.

An engine handles them well when it can read them. It reconciles your scheme pages against SEBI and AMFI disclosures, rating and research houses, investment platforms where schemes sit side by side, personal-finance desks and investor communities. Where those sources agree, the answer is confident and it cites somebody.

The question is whether it cites you. NAV, AUM, inception date, benchmark and riskometer published as text are readable. The same facts inside a factsheet image are not, and whoever restated them in a comparison table earns the citation instead. Your scheme is then described second-hand, on a page you do not control, next to funds you did not choose to sit beside.

Why does a returns figure go stale in an answer?

Because a return is a function of a date, and the engine is not reading your page at the moment it answers. It works from whenever it last crawled, or from training data older still. The number it shows an investor is usually a real number that was true once, which is a harder problem than an invented one: it looks correct, it does not match your factsheet, and nobody at your end sees it happen.

The second failure is quieter. A bare percentage with no window, no benchmark and no denominator cannot be checked, so a careful engine hedges it or drops it. Dropped reads to the investor as though your scheme has no performance worth mentioning. The fix is not more numbers. It is every number carrying its period and its benchmark in the same line, with trailing and rolling returns labelled as what they are.

The third failure is the one compliance will care about. An assistant that phrases past returns as an expectation has made a claim your firm cannot stand behind, in your name, to a buyer who will act on it. You cannot edit the model. You can make the correctly qualified version the easiest one to lift, and you can measure whether the answer still does it.

Cost is the line that survives between crawls

If returns are the fragile half, cost is the durable half, and it is under-published almost everywhere. Expense ratio, exit load and lock-in usually live on three different pages, or in three different documents, so an assistant assembling a cost picture has to stitch them together and will sometimes stitch them wrong.

Stated together, in text, they become the most quotable block on your scheme page. Cost is comparable by construction, which is exactly what a comparison answer needs, and it is still true next month. The thing your investor reaches for right after performance is also the thing whose wording you can actually control.

There is a second reason to publish the full cost. An investor who is convinced and cannot see how to begin is a lost SIP. Minimum amount, KYC status and the first step belong on the same page, because an answer that recommends you should also be able to say what happens next.

The five prompts, and what to read in the answer

Five prompts cover most of this category's research stage: best mutual funds to invest in India for long term; what is sip and how does it work; index funds vs active mutual funds; tax saving investment options; and a review of your own fund's performance. The first four are category questions where you compete to be named at all. The fifth is about you, and it is the one most likely to be answered from a source you have never read.

Presence is not one state, it is a ladder. The answer lists fund houses and yours is absent. Your AMC is named among providers with nothing attached. Your scheme page is cited as the source for a returns figure. Your fund sits inside the three-way comparison the investor is weighing. Your scheme is put forward as the better fit for this investor's horizon. Those are five different outcomes, and only the last one earns the SIP.

So read the citations, not only the mentions: which sources the engine built the answer from, which rival schemes appeared alongside yours, and whether your own page was used as evidence for anything. Then change one thing, publish the cost block as text and label every returns figure with its window, and run the same prompts again. One re-run tells you very little. Several runs on the same prompts tell you whether the source work moved the answer.

In Ansyra

Measure this for your own mutual funds brand

Ansyra runs the prompts mutual funds buyers actually type across every AI engine on your plan, records which sources each answer was built from, and shows where rivals are named instead of you.

Straight answers

Frequently asked

Why does an assistant quote a return that does not match our factsheet?
Because it is quoting whichever crawl it last read, and a return is only true for a date. The figure was usually real at some point, which makes it harder to spot than an outright error. Publish trailing and rolling returns as text with the window and the benchmark in the same line, so the version an engine lifts is the version that carries its own basis.
What should we publish so an engine states our cost correctly?
Expense ratio, exit load and lock-in on one page, in text, stated together rather than spread across three documents. Add the minimum investment amount and the KYC step. Cost is comparable and durable, so it is the part of a scheme page most likely to be quoted correctly and still be correct next month.
Does the fund manager's track record matter for AI visibility?
Less than most fund houses assume. SEBI's Investor Survey 2025 found the manager's track record named by far fewer mutual fund and ETF investors than either cost or performance, and an engine reflects that demand. Publish the manager detail, but do not expect it to carry an answer that your cost and performance facts are missing from.
How do we know whether AI answers name our schemes?
Run the prompts investors actually type and read what the answer cited. Record position, the sources behind each answer, and which rival schemes appear alongside yours, across ChatGPT, Gemini, Perplexity and AI Overviews. Re-check the same prompts after you change a page. One run is a snapshot, and a direction needs several.

See where you stand in AI answers

Run your real questions across ChatGPT, Gemini and Google's AI answers, free for 14 days. No connections required to start; all five AI engines on every plan.