AI Embryo Scoring and the Embryologist’s Assessment: How to Read Them Together

If your clinic mentions an AI score for your embryos, the natural question is whether it is more accurate than the person looking down the microscope. The short answer is that the two are not competing answers to the same question. An embryologist’s assessment describes what can be seen in the embryo’s structure and development. An AI score is usually a computer-generated number that ranks or estimates the likelihood of something, based on images it was trained on. Neither one is a diagnosis of the embryo, and neither guarantees a pregnancy.

This guide explains what each output represents, where uncertainty enters, what clinical evidence can and cannot currently show, and how to keep both pieces of information in view when you discuss embryo selection with your treating clinician.

Two different outputs, not two opinions on the same thing

Embryologists grade embryos using established morphological criteria: how the cells divide, how compact the cell mass looks, how expanded the cavity is, and similar visible features. This is a descriptive judgement made by a trained person, and it is usually recorded in a structured format so that it can be compared over time and between colleagues.

An AI system, by contrast, typically takes embryo images or time-lapse sequences and produces a score, a rank order, or a probability estimate. The output depends entirely on what the model was built to predict and on the data it learned from. A model trained to rank embryos by implantation potential is answering a different question from one trained to predict blastocyst formation, and the two numbers are not interchangeable.

That is why reading an AI score “alongside” a grade means asking what each one is actually telling you, rather than treating the higher number as the better embryo.

What an AI score usually represents

In most clinical settings, an AI output is best understood as a decision-support signal. It may:

• rank several embryos from the same cycle relative to each other;
• estimate a probability of reaching a developmental milestone or of implantation;
• flag images that look unusual compared with the training set.

None of these is a statement about the chromosomal makeup of the embryo. Morphology and genetic testing answer separate questions, and an image-based score does not substitute for a PGT result. If you are trying to understand how appearance and genetic testing can diverge, the explanation in embryo appearance versus genetic test results is a useful companion to this article.

Why the embryologist’s assessment still matters

An AI model sees only what it was given: images, frames, and the variables in its input. It does not know your age, your history, the stimulation protocol, or how the laboratory handled the sample that day. The embryologist does, and can weigh the score against the whole picture, including features the model was never trained to notice.

There is also a practical safeguard in having a human in the loop. A score that looks implausible, or an image of poor technical quality, can be recognised and set aside. The laboratory’s own procedures for how embryos are graded and selected are described in the embryologist’s role in Thai IVF clinics, which is worth reading if you want to understand who does what inside the lab.

Where the uncertainty comes from

Uncertainty around AI embryo scoring is not a reason to dismiss it, but it is a reason to ask precise questions. The main sources are:

• Training data. A model learns patterns from the embryos and outcomes in its development dataset. If that population differs from you or from the clinic using it, the score’s meaning may shift.
• Validation. Performance reported in one study does not automatically transfer to another clinic, another camera system, or another patient group.
• Laboratory calibration. Image quality, culture conditions and grading conventions vary between labs, and a model may need local adjustment before its output is meaningful.
• No universal standard. There is no single agreed scale for AI embryo scores, so a “0.8” from one system cannot be compared with a “0.8” from another.

This is the same kind of caution that applies to any test used in embryo selection. Regulators note that PGT-A itself has limitations and can produce incorrect findings, and that a result does not guarantee a successful pregnancy or a healthy child (HFEA). An AI score deserves at least the same scrutiny.

What the clinical evidence can and cannot show

Published research on AI in embryo selection is growing, but it is uneven. Studies often differ in what they predict, how they define success, and whether they compare the AI-assisted decision with what an experienced embryologist would have chosen anyway. A model that agrees closely with embryologists may be reassuring about consistency, but it does not by itself prove that using the score improves outcomes.

What the evidence generally does not yet establish is a universal claim that AI scoring is superior to morphological assessment for every patient, in every laboratory. Treat any clinic statement that presents AI as clearly better as something to examine: ask what it is compared with, in which population, and whether the clinic has its own data.

Reading a score and a grade together

A practical way to hold both in view is to ask three questions about any embryo under discussion:

1. What does the grade describe? Ask for the morphological grade in the laboratory’s own format, and what each element means.
2. What does the score predict? Ask what outcome the model was trained on, and whether the number is a rank within your cycle or an absolute probability.
3. How were they combined? Ask whether the score changed the ranking, confirmed it, or was noted but not used.

If the answers are vague, that is useful information in itself. A laboratory that can explain its own process is easier to work with than one that cannot.

Keeping a record you can discuss later

Embryo decisions are often made quickly, and the details are easy to lose. Before or during your cycle, ask whether you can receive the written morphological grading for each embryo, and, if an AI system is used, whether any AI report or score summary is available. Keep these with your other cycle records so that a later conversation, whether in Thailand or after you return home, starts from the same information.

If you are preparing questions for a consultation, questions about embryo quality covers the ground in more detail, and comparing embryo grading explanations between laboratories helps if you are weighing more than one clinic’s approach.

Who answers which question

Questions about what a score means for your own embryos, whether it should influence the transfer decision, and what the realistic chances are belong with your treating clinician and the embryology team. They can interpret the report in the context of your history. Coordination support, by contrast, covers the practical side: organising your records, requesting written grading or score summaries, arranging a remote consultation, and passing follow-up questions to the hospital. It does not interpret results or recommend a transfer decision.

If you are at the stage of gathering information before choosing where to be treated, a structured comparison of quotation inclusions and laboratory explanations is more useful than a ranking alone. The countrywide overview at researching IVF hospitals and clinics in Thailand is a neutral starting point for that comparison, not evidence that any provider produces better embryos or outcomes.

A short worked example

Suppose a cycle produces four blastocysts. The embryologist grades them, and the laboratory’s AI tool ranks them in a slightly different order. Rather than assuming the AI order is correct, you could ask: which embryo would have been chosen on morphology alone, which on the score, and why the team settled on the one it did. You could also ask whether the score was validated in this laboratory and what outcome it was trained to predict. Those three questions usually reveal whether the AI output is genuinely informing decisions or simply displayed alongside them.

The bottom line

An AI embryo score is a computer-generated estimate shaped by its training data and the laboratory using it. An embryologist’s assessment is a trained human description of visible embryo features. They can complement each other, but neither is a diagnosis, and neither guarantees an outcome. Read them together by asking what each one predicts, how they were combined, and what evidence supports the tool in that specific laboratory.

Frequently asked questions

Does a higher AI embryo score mean a better chance of pregnancy?

Not necessarily. An AI score is usually a ranking or probability estimate based on patterns in its training data, and it does not establish an individual embryo's chromosomal makeup or guarantee an outcome. Ask your treating clinician what the score was trained to predict and how it was used in your cycle.

Can I ask for the AI report and the embryologist's grading in writing?

You can ask the laboratory or your treating team whether written morphological grading and any AI score summary are available for your embryos. Keeping these with your cycle records makes later discussions easier, including after you return home.