What to Look For in an AI Portfolio Tool for eToro

Connecting any third-party tool to an eToro account is a decision about access, not just convenience. You are handing a piece of software the ability to see your positions and, depending on the mode you choose, to act on them. The right time to ask hard questions is before you connect, not after something goes wrong. This guide sets out the questions worth asking of any AI portfolio tool built for eToro, and uses Acubic's own answers as a worked example of what a complete answer looks like.

Does it publish how it actually builds a portfolio?

The first question is whether the construction method is documented anywhere you can read before connecting, rather than summarized in a sentence of marketing copy. A tool that will not show its work is asking you to trust an outcome without being able to check the process that produced it. Look for three things specifically: the optimization method used, the data categories that feed it, and a list of the model's own stated limitations. A tool that publishes features but not limitations has not actually published its method.

Acubic's methodology page sets out its default construction approach, Hierarchical Risk Parity, alongside the three-stage process behind it: screening the eligible universe, selecting from it, then allocating weights. It also lists five limitations the team considers material, including that historical performance does not guarantee future outcomes and that a universe built from currently listed instruments carries survivorship bias. A published limitation is not a marketing weakness. It is the sign that whoever built the tool understands what the model cannot do.

What happens when a rebalance is actually due?

Ask specifically what happens the moment a scheduled rebalance triggers, because this is where the difference between automated and unattended matters most. On a standard eToro connection, which runs one-click approval only, a scheduled rebalance prepares an order set and stops there. You review the proposed trades and approve or reject them before anything reaches eToro.

That is not a design choice a tool can override: any third-party application connecting to a standard eToro account is limited to one-click approval, so a tool that claims to trade unattended on your main account is describing something eToro itself does not permit.

There is one documented exception, and it is worth knowing about even if you never use it. Acubic can run without a per-order approval on the eToro mirrored account, a ring-fenced sub-account you fund separately from your main balance, where trades are mirrored in proportion to a ratio you set. That is the only eToro arrangement where a rebalance can complete without you clicking approve, and connecting your ordinary eToro account never turns that on by itself. The full mechanics of each mode are covered in one-click vs full-auto rebalancing.

Does the answer change on a different broker?

The approval question is not answered the same way twice, because the constraint comes from the broker's own account architecture, not from a single company's policy. A Trading 212 API key already covers the whole Invest account with no separate sub-account to isolate, so Acubic can run rebalancing on Trading 212 directly on the main account without a mirrored account or a per-order approval, if that is the mode you choose. eToro's standard connection has no equivalent: unattended execution there is limited to the mirrored account described above. A tool that gives you the same answer for every broker it supports has not actually looked at how each broker's API is built.

Does it treat an eToro account like an eToro account?

A generic portfolio tool built for no broker in particular will often miss details that matter specifically on eToro. Positions entered through copy trading are a good example: they sit in the same account as positions you built yourself, but a separate mechanism manages them, and a rebalancing tool that cannot tell the two apart will generate trades against holdings you never meant to touch.

Look for a tool that explicitly filters on where a position came from, so the managed set only ever includes positions you hold independently, and one that connects through eToro's own OAuth login rather than asking for your eToro password directly. On Acubic, authentication happens on eToro's own page, credentials never pass through Acubic's screens, and the resulting access token can be revoked from inside eToro at any time, independently of anything done in Acubic. The full credential flow is in how Acubic works with eToro.

Does it rebalance the gap, or rebuild the whole portfolio?

This distinction has a real cost attached to it in trading fees, so it is worth being precise rather than taking a tool's word for it. Rebalancing means comparing live positions against a target and trading only the difference; a position already at its target weight is left alone. Rebuilding means liquidating and repurchasing the full allocation regardless of how close it already is to target.

A tool that rebuilds on every cycle generates trading costs on positions that never needed to move. Ask specifically what happens when nothing has drifted: the honest answer, and the one Acubic gives, is that nothing trades at all in that case. The full construction logic behind the trigger and the trade set is covered in the rebalancing strategy model guide, and the difference between building a target and defending it against drift is covered separately in portfolio optimization vs rebalancing.

Is it explicit about what it will not do?

A tool that claims it can time the market or predict short-term price moves should be treated as a disqualifier rather than a feature, and the same is true of a tool that will not say plainly what it cannot do. Acubic does not predict short-term price moves, and no part of its construction method depends on forecasting which asset will outperform next. It does not promise or imply any level of return: the process is built to the risk constraints you set, not to a return target, and no track record is published, because publishing one would invite exactly the kind of promise this category should not make.

Ask the same question of any tool you are evaluating. A confident answer with no published method behind it is not evidence of anything, and a tool that will not say what it does not do is telling you something by omission.

Can you see what actually happened, not just what was proposed?

An automated system acting on a brokerage account is only trustworthy to the degree its actions are reconstructable afterwards. Brokers reject orders for ordinary reasons: an instrument becomes unavailable, a market is closed, a position falls below a minimum size. A tool that records only its successes is not giving you an accurate picture of what it did.

Acubic records every order submitted, filled, or rejected, along with the broker's own stated reason where one is given, and the remaining orders in a run are handled independently rather than the whole batch being abandoned when one fails. You can read that history after the fact for any past rebalance.

What happens if you leave?

Disconnecting a tool from a brokerage account should not be an event that forces a decision about your positions. If a tool liquidates on disconnect, staying connected quietly becomes a cost you did not agree to when you started. Ask directly what happens to open positions the moment access is revoked.

On Acubic, disconnecting does not liquidate anything. Whether you disconnect from inside Acubic or revoke access directly from within eToro's own settings, every position stays exactly where it is, and the only thing that stops is new orders being sent. The exit costs nothing beyond the decision to take it.

Applying this before you connect

• A published methodology, including the model's own stated limitations, not just its features.
• A visible approval step by default, with any unattended exception named specifically rather than implied.
• Broker-aware position handling, such as excluding copy-traded positions from the managed set.
• Minimal-trade rebalancing rather than a full rebuild on every cycle.
• An explicit statement of what the AI does not do, including short-term prediction and return promises.
• An audit log that records rejections, not only successful orders.
• A free exit that does not force a liquidation.

None of these questions are unique to Acubic, and a tool that answers all of them honestly does not have to be Acubic to be worth using. What they have in common is that each one can be checked against something the tool actually publishes, rather than taken on the strength of a landing page.

If you want to see how these pieces fit together end to end, how Acubic works walks through the full process from the first question to a finished, broker-connected strategy, and the eToro integration page has the setup steps, the access scope, and the revocation path in one place. If you would rather build a strategy first and decide on a broker connection later, the AI portfolio builder works the same way with no account connected.