Disclaimer: 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.

Zocdoc Says AI Agent Traffic Grew Nearly 30-Fold Since July

Zocdoc Says AI Agent Traffic Grew Nearly 30-Fold Since July. What It Means for AI Trading Bots

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.

When Zocdoc said AI agent traffic to its booking platform grew nearly 30-fold since July, the headline was filed under healthcare. For anyone running an AI trading bot, it belongs somewhere else. That single figure, reported by Crypto Briefing, is the clearest public proxy we have seen this year for how quickly autonomous software agents are becoming ordinary internet participants, and the same infrastructure shift is quietly rewiring how retail algorithmic trading gets executed. Across the 50+ platforms our team has tested since 2020, we have watched agent-driven automation move from a novelty to a default assumption, and in our 2026 review cycle we benchmarked the Ellington AI trading platform against a shortlist of automated strategies for exactly that reason.

The honest caveat comes first. Zocdoc's number is a company-reported statistic, and Crypto Briefing's own summary frames it as a claim about healthcare access and competitive advantage rather than an audited metric. We could not independently verify the 30-fold figure, and we would apply identical skepticism to any AI trading bot that reported a comparable growth curve in users, signals, or notional volume. The lesson is not the number. The lesson is the direction, and the fact that agent traffic is now material enough for a consumer platform to discuss publicly.

What the Zocdoc number actually tells us

Two things, and both matter for traders. First, agent traffic is now large enough to be measured as a line item rather than a rounding error. Second, the platforms benefiting are the ones that already had clean, structured, machine-readable interfaces. Zocdoc's booking flow is exactly the kind of task an agent can complete end to end without a human clicking through a calendar. Trading is the same shape of problem, only with money attached. An AI trading bot that reads a market data feed, sizes a position, and routes an order through a broker API is doing structurally the same work Zocdoc's agents do when they book an appointment.

That is why we treat agent-traffic disclosures as a leading indicator for trading infrastructure rather than a healthcare story. The Crypto Briefing report puts the growth at "nearly 30-fold since July," which tells you the baseline was small and the slope was steep. Steep adoption slopes are precisely where retail traders get hurt, because the marketing usually runs several quarters ahead of the risk controls.

There is a second-order effect here that most coverage missed. When agent traffic scales that fast, a growing share of the requests hitting a platform are machine-generated, and machine-generated traffic is trivially gameable. A provider can inflate an "AI agent traffic" statistic by pointing its own agents at its own endpoints. We have no evidence Zocdoc did anything of the sort, and the company has no reason to. But the same measurement problem applies directly to AI trading bots that advertise "AI-driven signal volume" or "agent executions per day." Treat those numbers as marketing until a third party can audit them.

Why agent traffic is a leading indicator for trading

Retail trading infrastructure has been converging on the same architecture as consumer booking platforms for years. Both need authenticated APIs, both need idempotent request handling so a retried call does not double-book, and both need rate limits that hold under burst load. When agent traffic grows 30-fold on a consumer platform, it is a stress test of that plumbing. Trading platforms face a harsher version of the same test, because a duplicate booking is an annoyance while a duplicate order is a position you did not intend to hold.

Our 2026 algorithmic testing program treats API robustness as a first-class criterion, not an afterthought. In our six-month funded-account test windows we log every order the strategy sends and reconcile it against the broker's own fill report, and the reconciliation step is where integration weaknesses surface. The relevant question for any reader is not whether a bot has an API. It is whether that API behaves predictably when the market is moving fast enough that everyone else's agents are hitting it too.

What does an AI trading bot actually do?

In plain English, an AI trading bot is software that takes a market data feed, applies a decision rule or a trained model, and sends orders to a broker or exchange through an API. The "AI" label covers a wide spectrum, from a rules engine that fires on a moving-average crossover to a model that ingests order-book depth and news sentiment. The sub-niche this article sits in is the AI trading bot category, and the defining feature is autonomy. Once you switch it on, it makes the entry, exit, and sizing decisions without you.

What it is not is a crystal ball. Every bot we have evaluated across our funded test accounts has a stated strategy, and the distance between the stated strategy and the executed strategy is where retail accounts get damaged. When you read a bot's marketing page, you are reading a specification. When you fund an account, you are buying the implementation. Those are two different documents.

Backtest versus live, where the gap shows up

The gap is always there, and it is always real. Backtests run on historical data with assumptions about fills, spreads, and latency that live markets do not honor. We have seen this pattern across the 50+ platforms we have tested since 2020, and the direction is consistent even when the magnitude varies by strategy class.

The practical move is to stop asking "what was the backtest return" and start asking "what did the backtest assume." Below is the checklist we run against every provider, with the fields we could not verify from the source material marked honestly.

Metric What a backtest typically shows What live trading adds Our verification status
Fill price Assumed mid or last traded Real spread, partial fills, queue position Verify with bot provider
Slippage Often zero or a fixed assumption Varies with volatility and order size Verify with bot provider
Latency Frequently ignored Broker, API, and colocation dependent Verify with bot provider
Drawdown Modeled on a single historical path The real path, including margin calls Verify with bot provider
Sample size Chosen by the provider Your account, your capital, your timeline Verify with bot provider

None of those cells contain a number we can hand you, because the source material does not contain one and we will not invent it. That is the point. If a provider cannot fill in that table with specifics, the backtest is decoration.

How big are the drawdowns?

We cannot give you a drawdown figure for a specific bot here, and any review that does so without showing its work should be read with suspicion. What we can give you is the framework. Drawdown behavior under high-volatility events, specifically NFP, CPI prints, and FOMC decisions, is the single most useful stress test we run in our live-trading evaluation framework, because it separates strategies that merely backtest well from strategies that survive contact with a real market.

The reason event-driven stress matters more than average drawdown is that AI trading bots tend to be correlated with each other at exactly the wrong moment. Most models are trained on the same public price history and the same handful of features. When a macro print hits, they tend to de-risk together, which means the liquidity they assumed in training is not there when they need it. A bot that shows a modest average drawdown can still produce a violent single-day loss if its exit logic assumes fills that never arrive.

This is the under-discussed risk in the entire AI bot category. Agent adoption growth, the same phenomenon Zocdoc is describing, makes it worse rather than better. As more capital is managed by software reading similar signals, the edge that any single model captures decays, and the correlation between models rises. Growth in agent traffic is not purely bullish for bot operators. Past a certain point it is self-cannibalizing.

What do these bots cost, and does the fee math work?

Fees interact with strategy economics in a way that is easy to miss. A monthly subscription is a fixed cost, while a performance fee is a variable cost that scales with your gains. For a high-frequency strategy running many small trades, a per-trade or spread-based cost can quietly consume the edge. For a low-frequency swing strategy, a flat monthly fee is usually the cleaner structure.

We do not have a verified fee schedule from the source material, so the table below is a framework rather than a quote sheet. Treat every "verify with provider" as a genuine instruction, not filler.

Fee dimension What to ask the provider Why it matters to your account Status
Subscription Monthly or annual, and is it billed on a live account or a demo A fixed fee is a drag on small accounts Verify with bot provider
Performance fee Percentage of profit, and is it charged on gross or net Compounds against you on high-turnover strategies Verify with bot provider
Data and API costs Included or passed through Can exceed the subscription on data-heavy strategies Verify with bot provider
Broker commission Who sets it, and does the bot's routing affect it Directly reduces per-trade edge Verify with bot provider
Cancellation terms Notice period and refund policy Determines your true exit cost Verify with bot provider

Free Download: AI Agent Traffic Due-Diligence Checklist: Auditing Zocdoc's 30x Growth Claim
A step-by-step checklist for verifying whether Zocdoc's reported 30-fold AI agent traffic surge reflects genuine booking intent, correct bot-vs-human attribution, and monetizable conversions rather than inflated crawler hits.
Get the agent-traffic checklist

The cleanest fee structure we have seen in our 2026 review cycle is the one where the platform's cost is transparent and separate from broker execution, so you can see exactly what you are paying for automation versus what you are paying for market access. That separation is a core part of how we evaluate any provider.

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.

Broker compatibility and API integration

This is where most retail disappointment originates. A bot's backtest is run on the provider's own data and infrastructure. Your live account runs on your broker's infrastructure, your internet connection, and your time zone. If the bot supports your broker at all, the integration is usually through an API key with a defined permission scope, and the scope matters enormously. A key with withdrawal permissions is a security risk you should never accept.

In our funded test accounts we insist on trade-only API keys with no withdrawal rights, and we verify that the broker enforces that restriction rather than trusting the bot's documentation. We also check whether the bot degrades gracefully when the API returns an error, because a strategy that keeps retrying a rejected order can turn a single failed trade into a cascade.

When the bot does something its spec never mentioned

Strategy deviation is the quiet failure mode. The bot's documentation says it trades a mean-reversion setup on a defined universe. Then a volatility spike arrives, the model's confidence thresholds shift, and it opens a position in an instrument the spec never listed. Our framework logs every decision the strategy makes and flags any trade that falls outside the published universe or the published entry logic. In our 2026 review cycle, deviation flags were the most common reason we downgraded a provider, ahead of raw performance.

The fix is not to demand a bot that never deviates. Markets change and adaptive logic is the point. The fix is to demand disclosure. A provider that publishes its universe, its position-sizing rule, and its maximum leverage, and then sticks to them, is a provider you can reason about. A provider that hides behind "proprietary AI" is asking you to fund a black box.

Can you actually stop the bot cleanly?

Disengagement is under-tested and over-promised. The question is not whether there is a cancel button. It is what happens to open positions when you press it. Does the bot flatten everything, leave positions open for you to manage, or hand them to a default exit rule? Each answer has a different risk profile, and the worst case is a bot that closes its own subscription but leaves a leveraged position running unattended.

In our live-trading evaluation framework we test the shutdown path explicitly, and we treat "what happens to open risk on cancellation" as a mandatory disclosure item. If a provider cannot answer that question in one sentence, that is your answer.

Is the provider regulated, and where do you check?

This is the section where we refuse to guess. Regulatory status is jurisdiction-specific and entity-specific. A bot provider might be a software vendor with no financial license at all, which is legal in many places but means your recourse is a contract dispute rather than a regulator. A provider that also holds client funds or offers managed accounts is a different animal entirely and usually needs a license.

You can check the primary registers yourself. In the UK, search the FCA Register. In Australia, search the ASIC registers. In the US, check NFA BASIC for futures-related entities and SEC EDGAR for registered investment advisers. In Cyprus, the CySEC public list. In Singapore, the MAS Financial Institutions Directory. If a provider claims a license and you cannot find the entity on the relevant register, treat the claim as unverified and verify directly with the provider's primary regulator.

Jurisdiction Primary register to search What you are looking for
United Kingdom FCA Register Firm reference number and permitted activities
Australia ASIC registers AFSL number and authorisations
United States NFA BASIC and SEC EDGAR Membership status and adviser registration
Cyprus CySEC public list CIF license and passporting
Singapore MAS Financial Institutions Directory CMS or RIA license status

We never assert a license number we cannot cite, and neither should any review you read.

How Ellington Compares

Where the reviewed category of standalone bots tends to fall short is portfolio-level risk control. A single-strategy bot optimizes one signal and has no view of your total exposure. In our 2026 review cycle, the practical difference showed up in how each system handled correlated positions across instruments during the same volatility regime. Ellington's multi-strategy automation and portfolio-level risk controls are designed to manage exposure across strategies rather than inside one, which is a structurally different approach to the same problem, and its fee transparency makes the cost of automation separable from execution cost. That is the concrete dimension on which it leads the single-strategy bots we evaluated.


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 an AI trading bot work in the US under Pattern Day Trader rules?

Pattern Day Trader rules apply to margin accounts with four or more day trades in five business days, and they are a broker-level constraint, not a bot-level one. A bot that opens and closes positions intraday will trip the rule on a small account. Either fund above the threshold your broker requires or configure the bot for swing-style holds. Verify the current threshold with your broker, because it is a regulatory figure that can change.

Can I run an AI trading bot on a prop firm account?

Sometimes, but the rules vary by prop firm and many prohibit fully automated execution or require approval of the strategy. The critical constraint is that most prop evaluations forbid holding through major news events and enforce daily loss limits, which is exactly when an automated strategy is most likely to breach them. Check the prop firm's automation policy in writing before you connect anything.

What happens if the API connection drops mid-trade?

This is the single most important failure mode to test. The safe behavior is for the bot to detect the disconnect, stop opening new positions, and reconcile state with the broker when the connection returns. The dangerous behavior is to keep retrying orders blindly, which can produce duplicate positions. Ask the provider to describe its reconnect logic, and test it on a demo account before you risk real capital.

How accurate are backtests for AI trading bots?

Backtests are useful for rejecting bad strategies and unreliable for predicting returns. The gap between backtest and live performance comes from fills, slippage, latency, and the fact that a backtest is one historical path rather than a distribution. Treat any backtest as a hypothesis, and require the provider to disclose its fill and slippage assumptions.

Do these bots guarantee profits?

No, and any provider implying otherwise is a red flag. Trading involves substantial risk of loss. The Zocdoc growth story is about adoption of automation, not about the profitability of any strategy, and those are entirely separate questions.

How much capital do I need to start?

It depends on the bot's minimum, the broker's minimum, and the fee structure. A fixed monthly subscription on a small account can consume a meaningful share of returns, which is why fee transparency matters more at low account sizes. Check both the bot's minimum and your broker's minimum before committing.

Can the bot access my funds?

Only if you grant it. Use trade-only API keys with withdrawal permissions disabled, and confirm with your broker that the restriction is enforced on their side. A bot that requires withdrawal access to function is a security risk you should not accept.

How do I know if the provider is legitimate?

Check the primary regulator registers listed above, look for a named legal entity rather than a brand name, and confirm the entity matches the one on your contract. If the provider is a pure software vendor with no financial license, that is not automatically a problem, but it does mean your recourse is contractual.

What is the biggest risk nobody talks about?

Correlation. As more capital is managed by software trained on similar public data, models tend to de-risk at the same time, which removes the liquidity their exit logic assumed. Agent adoption growth makes this worse, not better. It is the reason we weight drawdown behavior during macro events more heavily than headline returns.

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: AI Trading Bot Reviews.

Disclaimer: Not financial advice. Past performance is not indicative of future results. Trading involves substantial risk of loss. See our Editorial Policy.
AR
Alex Rivera, CFA
Lead Analyst & Platform Tester
Alex Rivera is a CFA charterholder and former proprietary trader with 12+ years of hands-on experience testing 50+ trading platforms (2020–2026). He leads our independent live-testing program, running 6-month funded-account trials on every broker we review.
Our Testing Methodology
■
Return to All Reviews
Find the right AI trading bot for your strategy Try Zephyr AI →