AI stock screeners in India: inspect the query before the list

Which part of the screener uses AI?

There are separate jobs here: translating a sentence, selecting database fields, running filters, and explaining the returned companies. A tool may automate one job while leaving the others to you. Treating all four as the same capability makes comparisons misleading.

Trendlyne's mode documentation explicitly describes plain-English query creation and manual editing. Its MarketMind access guide describes another AI feature reached through several dashboards. StockEdge describes custom conditions and automatic chart recognition. Its AI article also puts the AI label on features it describes as rules: Sector Rotation ranks sectors "based on technical indicators like relative strength, momentum, and trend" and is called "a classic example of financial AI at work", and its trading strategies are "predefined combinations of scans". The article does not say how a model is involved in either. Screener's current pricing page lists both Screener AI and AI summaries. Those descriptions establish features, not their accuracy on your question.

Start with the stock screener guide if you need the difference between a financial filter and a technical scan. Here the useful question is whether an assistant preserves the conditions you intended.

What the primary pages establish

Documentation opened 14 October 2026. No accounts were benchmarked and no stock selections are shown.

Tool	Documented job	Evidence limit
Trendlyne	AI Mode builds a screener query from plain English; query can be edited.	MarketMind access locations do not establish identical behaviour or plan entitlement.
StockEdge	Custom scans combine conditions; automatic AI chart patterns appear in Pro and Club.	Its AI article labels rule-based features (sector rankings, predefined scan combinations) as AI and does not describe a natural-language scan.
Screener.in	Free stock screens and custom ratios; pricing also lists Screener AI and AI summaries.	Credits and summaries are distinct from proof that every requested screen can be generated.
Perplexity	India launch (14 Aug 2025) covered BSE/NSE prices, financials and historical downloads; a 4 Sep 2025 changelog says its stock screener is now available for Indian markets.	Perplexity's pages do not show how the screener turns a question into filters; we did not test it.
ChatGPT	OpenAI documents working with a supplied spreadsheet or CSV and reviewing a cleaned copy.	This does not establish an exhaustive Indian exchange database or a live screener entitlement.

Rows follow sources modes/marketmind, edge/edgeplans, screener/query, perplexity/perplexity-sep and openai respectively. Documentation checked 14 Oct 2026. Availability in an individual account remains a separate check.

Sources: [1], [2], [3], [4], [5], [6], [7], [8], [9]

Five questions with exact meanings

A comparison needs a specification before it needs screenshots. The following prompts are an illustrative test pack, not recommended investment filters. They use fictional rows, so no current or historical security is suggested. The numbers are deliberately chosen to expose boundary mistakes.

Debt and ownership: Find rows with debt exactly zero, market capitalisation from ₹5,000 crore through ₹15,000 crore inclusive, and promoter holding strictly higher than the previous quarter. This avoids treating a vague word such as midcap as an agreed classification. It also separates zero debt from a low debt-to-equity ratio.

Profitability and valuation: Require ROE of at least 15%, together with a positive P/E no greater than 20. A negative P/E must fail. An assistant that simply interprets low P/E as anything below the ceiling can accidentally include a loss-making row.

Cash conversion: Require operating cash flow strictly greater than net profit, with net profit positive, using the same assumed annual period and units. A result based on different reporting periods cannot answer this question even if the arithmetic is correct.

Trend field: Require the supplied close to be strictly above the supplied simple moving average. Equality must fail. This is a comparison of given columns, not a request to calculate a moving average from an unspecified history.

Volume field: Require supplied volume to be at least 1.5 times the supplied average volume. Equality must pass. The illustrative average excludes the observation being tested; changing that convention would change the test.

Fictional financial input sheet

Row	Cap ₹ crore	Debt ₹ crore	Holding: now / prior %	ROE % / P/E	Cash flow / profit, assumed units
A	8000	0	55 / 54	18 / 18	120 / 100
B	10000	0	54 / 55	20 / 22	80 / 100
C	6000	25	60 / 58	16 / 15	150 / 100
D	9000	0	40 / 39	12 / 10	70 / 80
E	2500	0	65 / 60	25 / -5	30 / -20
F	20000	0	51 / 50	15 / 20	100 / 100

All values are author assumptions, not company data. Each column is set on its own to test one rule, so columns are not meant to agree with each other: for example, E's positive ROE beside a loss, or a P/E that doesn't equal cap divided by profit, could not both happen in real accounts. Sources explain the screen-building workflows; no vendor supplied these rows.

Sources: [6], [1]

Fictional price and volume input sheet

Row	Close / SMA, assumed units	Volume / average, assumed units
A	210 / 200	160 / 100
B	190 / 200	200 / 100
C	220 / 200	100 / 100
D	200 / 200	150 / 100
E	120 / 100	151 / 100
F	110 / 100	140 / 100

Synthetic inputs belong to the same exercise. No ticker, trade date, signal or market price is implied.

Sources: [1], [3]

Read the answer key by exclusion

The locally calculated matches are A and D for debt/ownership; A, C and F for profitability/valuation; A and C for cash conversion; A, C, E and F for the close/SMA test; and A, B, D and E for volume. These are set comparisons, so order does not matter.

Inspect the rows that almost qualify. F satisfies the ROE boundary at 15% and the P/E boundary at 20, but its assumed capitalisation is outside the first screen's range. E has a negative P/E, so it fails the second screen despite a high assumed ROE. D sits exactly on its moving average and therefore fails a strictly-above test. D's volume is exactly 150 against an average of 100, so it passes the inclusive volume test.

For the cash-flow screen, F's 100 equals its profit of 100 and must fail the strict inequality. E's cash flow exceeds its negative profit, but the additional positive-profit condition excludes it. This is why it helps to show the complete Boolean expression rather than accept an attractive explanation of the result.

The exercise proves only that the answer key follows its declared inputs. Real Indian financial datasets add reporting dates, consolidated versus standalone accounts, missing values and revised filings. None of those complications has been simulated here.

A vendor comparison that can be reproduced

1. 1

   Freeze the research universe

   Use the same dated, licensed dataset for every run. For an educational market example on these AI pages, use observations at least thirty days old. Preserve identifiers and the original file so changes in company names or symbols cannot silently change the sample.

2. 2

   Record the translation separately

   Save the prompt, generated query, field names and exclusions before reading the stock list. A correct list from an altered question is not a pass. If the tool exposes only prose, mark query inspection unavailable instead of assuming its hidden method.

3. 3

   Verify the result outside the assistant

   Evaluate each condition from the preserved sheet. Count both incorrect inclusions and omitted qualifying rows. Record untestable fields separately; missing access is neither a correct result nor an incorrect one.

4. 4

   Keep edits visible

   Run the original prompt before coaching the tool. Preserve any correction and repeat with the same data. A manually repaired query may be useful, but it is a different result from an accurate first response.

Where the free plan stops

Free access can mean free screening, free summaries, or limited credits. Screener's current comparison lists free stock screens and custom ratios, ten AI summaries per month, and Premium at ₹4,999 per year including GST. The separately labelled Screener AI row gives Premium credits worth ₹500. It does not say those credits buy unlimited AI use.

StockEdge's comparison lists two custom scans in Premium and fifty in Pro and Club; automatic AI chart patterns are shown in the latter plans. The same grid lists a "Stock AI" item under ready stock ideas, ticked for Club only, without saying what it does. Its pricing page contains promotional and billing presentations, so this guide does not turn one displayed monthly equivalent into a guaranteed annual invoice. Confirm the final amount and tax treatment for the actual purchase.

Trendlyne's access article tells you where to find MarketMind. It does not, by itself, settle your plan's allowance or the charge for every AI action. We have left that price unquoted. Perplexity's changelog of 4 September 2025 says its stock screener now covers Indian markets, but its plan pages don't say whether screening counts against the free plan's limits. Our Perplexity Finance review separates those claims in more detail.

Reject a result when its definitions moved

• The period changed

  Latest quarter, trailing twelve months and latest financial year are different inputs. Keep the requested period beside each column.

• An unknown became zero

  Missing debt or profit is not proof of zero debt or break-even earnings. Require an explicit missing-data policy.

• A percentage became a percentage-point change

  Promoter holding moving from 50% to 51% is a one-percentage-point increase. Do not let the explanation silently substitute a relative percentage calculation.

• The universe became a web sample

  A collection of cited companies can be useful research, but it is not evidence that every eligible company was tested.

Choose a workflow you can audit

If your main need is expressing a financial condition, inspect a generated query alongside a manual query. If your main need is comparing company filings, verify the underlying periods and footnotes before adding AI summaries. If you need price scans and alerts, the intraday screener comparison considers refresh and delivery separately.

A general assistant working from your spreadsheet can help make a small exercise transparent: preserve row IDs, ask for the included and excluded sets, and request the reason for each exclusion. OpenAI's file-cleaning guidance supports reviewing a separate cleaned output and a data-quality note. The completeness and licensing of the Indian market data still need their own evidence.

For financial research versus technical screening, Chartink and Screener.in have a more focused comparison. A useful choice follows the job and the evidence available in your account. A fluent explanation, an AI badge and a long returned list do not measure whether the original conditions survived.

Sources

Every figure on this page comes from one of these. Each was checked on the date shown.

1. [1]Trendlyne: Easy, AI and Manual screener modes · checked 14 Oct 2026
2. [2]Trendlyne: accessing MarketMind AI · checked 14 Oct 2026
3. [3]StockEdge blog: How StockEdge uses AI in finance (16 Jan 2026, updated 11 May 2026) · checked 14 Oct 2026
4. [4]StockEdge: feature and plan comparison · checked 14 Oct 2026
5. [5]Screener: current free and Premium comparison · checked 14 Oct 2026
6. [6]Screener: creating screens · checked 14 Oct 2026
7. [7]Perplexity: India Finance launch · checked 14 Oct 2026
8. [8]Perplexity changelog: tools for Indian stocks, 4 September 2025 · checked 14 Oct 2026
9. [9]OpenAI: cleaning and reviewing supplied tabular data · checked 14 Oct 2026