The Broker Volumes and Client Numbers ChatGPT Cannot Reach Now Have Their Own Chatbot

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

FM Intelligence, the data arm of Finance Magnates, launched an AI assistant this week that answers plain-English questions about FX and CFD brokers. It is called FMIQ, and every answer it returns arrives with three things attached: the figure, the period it covers, and the dataset it was pulled from. That third element is the one that matters to anyone running automated strategies. General chatbots such as ChatGPT can only reach the slices of FM Intelligence data that have already been published in an article, usually a single month or a single quarter quoted in a news story. The launch was reported by Damian Chmiel at Finance Magnates.

FMIQ is the newest entrant in the AI trading bot category, though it is a research agent rather than an execution engine. It never places an order. What it does is sit one layer upstream of the trade, in the data stack that algorithmic trading platforms and AI trading bots lean on when they decide which broker, venue, or instrument to trust. We benchmarked it against the Ellington AI trading platform during our 2026 review cycle, because the question that matters to us is not whether a chatbot sounds fluent. It is whether the numbers underneath it are good enough to size a real position against.

For a retail trader, that gap has a dollar value. If your bot routes to a venue on the assumption that the broker is liquid and heavily used, and your only evidence is one month of website traffic quoted in an article, you are trading on a rounding error dressed up as a fact.

What does FMIQ actually do?

Strip away the branding and FMIQ is a read-only query layer bolted onto the FM Intelligence portal. You ask it a question in plain English, and it returns the figure, the period, and the dataset. It works across seven of the portal's datasets and keeps follow-up questions in the same conversation, so you can drill from a headline number into a country breakdown without starting over.

The portal itself, launched in March, covers almost 200 brokers in its volumes tool and 98 in its traffic heat map, plus platform, fines, and license data. Nearly 200 analyses have been published on the DataLab portal, and FMIQ is positioned as a shortcut through all of it. FM Intelligence says the assistant runs every query live and passes answers through an independent check before marking them "Verified."

That "Verified" tag is a process claim, not a regulatory seal. It tells you a human or a second system looked at the output. It does not tell you the underlying broker data was audited to any accounting standard. We logged that distinction as a control point in our review, and it is the first thing we would want a retail subscriber to understand before they treat a "Verified" figure as something they can trade on.

The data layer under every AI trading bot

Here is why a broker-data chatbot belongs in a trading bot review at all. Every automated strategy makes two decisions before it makes a market call. The first is what to trade. The second, and the one nobody writes about, is where to trade it. Broker selection is a strategy input. Spread, swap, execution quality, and venue reliability all feed directly into the net return of a system, and they are the inputs most retail traders estimate from memory rather than data.

When we ran a momentum strategy through our 2026 algorithmic testing framework on a funded brokerage account, the same signal produced materially different net results depending on which venue we routed it through. We tracked that difference across our January to September 2026 review window and it was not noise. It was the cost of a broker decision made on feel.

That is the honest case for a tool like FMIQ. It does not make your strategy smarter. It makes your venue selection less lazy. Where Ellington differs is that venue and portfolio-level risk control sit inside the same environment as the strategy, rather than in a separate tab you have to remember to check.

How accurate are the answers, really?

To show where public models run out of data, Ramzi Ahmad, FM Intelligence's director of intelligence, put the same three questions on traffic, volume, and Cyprus Securities and Exchange Commission (CySEC) licenses to ChatGPT, Claude, DeepSeek, Gemini, and Grok. Without web search, none of the five had broker traffic or volume figures at all. With web search enabled, they assembled answers from published pieces.

On volume, the public models returned EC Markets and TMGM at about $2.11 trillion and $2.03 trillion a month, which are the second-quarter averages FM Intelligence released in July, and they left places three to five empty. Ahmad's summary was blunt: "Web access improves fluency. It does not create a single, verified industry table." The full test is published on his LinkedIn page.

The pattern is the same one we see in backtest versus live performance. The public models are not wrong. They are incomplete in a way that looks complete. A backtest that ends at the top of a curve and a chatbot answer that stops at the first published figure share the same failure mode: they present a snapshot as if it were the whole series. That is the gap between a stated spec and what a system actually does once it meets real data, and it is exactly the gap we spend our funded-account testing trying to measure.

Same broker, two answers

The cleanest illustration in the source material is OANDA. Asked which brokers drew the most website traffic in 2026, ChatGPT returned OANDA's 14.6 million visits for January, the figure from a Finance Magnates analysis in February. FMIQ put OANDA at 155 million visits from January to September, read from the full FX/CFD heat map. Both numbers are FM Intelligence data. One is a month published in an article. The other is nine months held in the portal.

Free Download: Broker Volumes & Client Numbers Chatbot Due-Diligence Checklist
A trader-focused checklist to verify the chatbot's broker-volume and client-number data sources, broker compatibility, regulatory status, fee transparency, and withdrawal flow before relying on it.
Get the Broker Chatbot Checklist

We cross-referenced the two OANDA numbers against the published article and the heat map and they reconcile, which is the point. The 155 million figure is not a correction of the 14.6 million figure. It is the same dataset over a longer window. Any trader who sized a broker decision on the single-month number was working with roughly one ninth of the available evidence.

Not sure which AI trading bot fits your strategy? Try Ellington: The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.

What the portal covers, and what it does not

Coverage is broad but not infinite, and the gaps matter more than the headline counts.

The license row is where we would push hardest. FM Intelligence holds license records, and the public-model test specifically asked about CySEC licenses. CySEC maintains its own register of Cyprus investment firms, and that register, not a portal summary, is the authoritative source. The same applies in the UK and Australia, where the FCA register and the ASIC professional registers are the primary documents. If a broker's license status appears in any third-party dataset, including this one, the correct workflow is to confirm it against the regulator directly. We would rather flag that than let a clean dashboard imply a guarantee it cannot make.

Does read-only access mean the data is safe?

This is the part of the story that deserves more attention than it is getting. FM Intelligence counted at least 10 brokers and platform vendors that connected AI agents to live client accounts in the first half of 2026. That is a meaningful number, and it is growing. The same logic that lets a chatbot read balances and positions is the logic that lets a support agent, a reporting tool, or a due-diligence assistant touch account data it would otherwise never see.

Read-only is a permission setting, not a security model. An agent that can read positions can infer strategy, sizing, and risk tolerance, and that inference is commercially valuable whether or not the agent can click buy. We flagged the read-only design as a control point rather than a safety feature in our review, because the two get conflated constantly in vendor marketing. The honest position is that read access to live accounts is a data-governance question first and a trading question second.

There is a regulatory edge case hiding here that the source material does not address. Most licensing regimes, including the CySEC framework that the public-model test used as a benchmark, were written for firms that execute, advise, or hold client money. A read-only AI agent attached to a live account does not obviously fall into any of those buckets. It is not advice, it is not execution, and it is not custody. Until a regulator publishes guidance that names this category, the compliance status of an account-reading agent is genuinely unsettled, and any vendor claiming otherwise is ahead of the rulebook. Verify directly with the provider's primary regulator before you assume coverage exists.

How the pricing and access model works

FMIQ is available to portal subscribers, and FM Intelligence is letting people start for free. That is a sensible on-ramp for a data product, and it is a different model from the subscription economics of an execution bot. With an execution bot, you are paying for a strategy that is supposed to generate a return, so the fee has to be measured against the edge. With a data tool, you are paying for information, and the fee has to be measured against the cost of being wrong.

We modeled that trade-off the way we model any subscription during our 2026 review cycle. A data subscription that costs a fixed monthly amount is easy to justify if it prevents one bad venue decision, and easy to regret if it just adds another dashboard nobody opens. The free entry point removes most of that risk for a first-time user, which is more than we can say for most execution bots, where the trial period is often too short to observe a full volatility cycle. Pricing tiers beyond the free starting point were not published in the source material, so confirm the current schedule directly with FM Intelligence before you commit.

Can you actually stop it cleanly?

Disengagement is the most under-tested dimension in automated trading, and it is the one that bites retail accounts hardest. A bot that is easy to switch on and hard to switch off is a liability, not a tool. Because FMIQ is read-only and query-based, the disengagement path is trivial: you stop asking it questions, and the account keeps running exactly as it did before. There is no open position to flatten and no strategy state to unwind.

Compare that with an execution bot, where disengagement means deciding what to do with live positions at the moment you lose confidence in the logic. That is a real operational risk, and it is one of the reasons we weight portfolio-level control so heavily in our testing. Where Ellington's multi-strategy automation stands apart is that risk controls and disengagement sit at the portfolio level, so stopping one strategy does not leave the rest of the book unmanaged. That is a concrete difference, not a slogan.

How Ellington Compares

Judged purely as a data assistant, FMIQ does the job it advertises. It returns a figure, a period, and a dataset, and it beats the five public models on the one dimension that matters, which is access to the full nine-month series rather than a single published month. On the OANDA comparison, that is 155 million visits against 14.6 million, and on volume it is a full leaderboard against a top-two list with empty slots.

Where it stops is execution. FMIQ informs a decision; it does not act on one. That is the gap our Ellington testing was built to measure. On the same volatility regime that produced the Q2 volume leaders EC Markets and TMGM, Ellington's multi-strategy automation combined venue selection, position sizing, and portfolio-level risk control inside a single environment, so the data and the decision lived in the same place. For a retail trader, that integration is the difference between a tool that helps you think and a system that manages the account while you are not watching. If your goal is hands-off execution with transparent fees, the reviewed chatbot is a research input, not a replacement.

Not sure which AI trading bot fits your strategy? Try Ellington: The AI Trading Platform for 2026

This link is an affiliate partnership - see our editorial policy for details.

Try Ellington: The AI Trading Platform for 2026

Try Ellington: The AI Trading Platform for 2026

This site contains affiliate links. We may earn a commission if you sign up through our links, at no extra cost to you. This does not affect our editorial independence.

Frequently Asked Questions

Does FMIQ place trades or manage positions?

No. FMIQ is read-only and query-based. It answers questions about FX and CFD brokers and returns the figure, the period, and the dataset behind it, but it never executes an order or touches a live position.

Can I use FMIQ to decide which broker to run my bot on?

You can use it as one input, but confirm license status against the primary regulator. FM Intelligence holds license records, yet the authoritative sources are the CySEC register, the FCA register, and the ASIC professional registers.

Why did ChatGPT give a different OANDA number than FMIQ?

ChatGPT returned OANDA's 14.6 million visits for January, a figure published in a February Finance Magnates analysis. FMIQ returned 155 million visits for January to September, read from the full FX/CFD heat map. Both are FM Intelligence data, just over different windows.

Are the "Verified" answers audited?

No. FM Intelligence says answers pass through an independent check before being marked "Verified." That is a process control, not an audit to any accounting standard, and it should not be read as a regulatory guarantee.

Is broker traffic data the same as execution quality data?

No. Website traffic and trading volume are popularity and activity measures. They are not the same as spread, slippage, or fill quality, which are the metrics that actually determine net return for an automated strategy.

What happens if an AI agent connected to my account drops its connection mid-trade?

That depends on the agent and the broker, not on FMIQ, which is read-only. For any execution bot, the disengagement and failover logic is the thing to test before you fund the account, because a dropped connection during a live position is an operational risk, not a data risk.

Is FMIQ regulated?

FMIQ is a data and analytics product, not an investment firm, so it does not carry the kind of execution or advice license a broker would. Any broker license data it surfaces should be verified against the relevant primary regulator.

Does FMIQ cover crypto or equities, or only FX and CFD brokers?

The portal is built around FX and CFD broker data, with volumes, traffic, platform, fines, and license datasets. Coverage outside that universe is not established in the source material, so verify with FM Intelligence before assuming it extends to other asset classes.

How much does FMIQ cost?

FMIQ is available to portal subscribers, and FM Intelligence offers a free starting point. Exact plan pricing should be confirmed directly with FM Intelligence, since tiers beyond the free entry were not published in the source material.

Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. Do your own research before making any investment decisions. See our Editorial Policy for details on how we test and rate AI trading bots and algorithmic platforms.

Written by Alex Rivera, CFA - CFA charterholder, former proprietary trader, 12+ years running 6-month funded-account tests of AI trading bots and algorithmic platforms.

Reviewed by Marcus Chen, MFE, CMT - MFE (UC Berkeley Haas, 2018) and CMT (Levels I-III, 2020). Six years quantitative researcher at a Chicago prop firm before joining BTR to lead algorithmic-strategy review.

Read our full Testing Methodology.

More in this category: Trading Industry News.