Designer babies

Designer babies is a popular term for children whose genetic characteristics have been deliberately chosen by their parents. It covers two very different things. One is Embryo selection, which is routine in fertility clinics and can reliably deliver only a short list of outcomes. The other is heritable Human germline editing, which has been attempted once, illegally, and would face a different and harder set of technical obstacles. Most public argument treats the two as a single prospect, which is where the confusion begins.

§Origins of the term

The phrase is journalistic rather than technical, built on the same pattern as "designer drugs" and in wide circulation by the 1990s. It attached itself to a sequence of real clinical developments. In vitro fertilisation produced its first birth in 1978; preimplantation genetic diagnosis followed in 1990, when a British team biopsied embryos and selected female ones for couples at risk of X-linked disease.¹ The film Gattaca fixed the popular image in 1997, some years before any of the genomic tools that would be needed to realise it existed.

A defining case arrived in 2000, when a couple used preimplantation diagnosis both to avoid Fanconi anaemia and to select an embryo whose tissue type matched their affected daughter's, so that cord blood from the resulting child could treat her.² The "saviour sibling" framing dominated coverage for a decade and set the pattern for subsequent debate: a specific, medically motivated use, discussed as though it were the first step of a general capability.

§What selection can actually deliver

Selection is constrained by what the parents already carry. An IVF cycle produces a set of embryos that are recombinations of two existing genomes; no embryo can have an allele neither parent possesses. Within that set, selection is highly reliable for categorical outcomes determined by one locus or by chromosome count.

• Single-gene disorders. Where both parents carry a recessive allele, roughly a quarter of embryos are affected and can be identified with high accuracy. This is PGT-M, and it works.
• Chromosomal aneuploidy. PGT-A screens for the wrong number of chromosomes. Its value in improving live-birth rates remains debated, but the measurement itself is dependable.
• Sex. Sex is a categorical genetic fact, which is why it is the one non-medical trait selection delivers with near certainty, and why most jurisdictions that regulate embryo testing address it explicitly.
• Tissue type. HLA matching, as in the saviour-sibling case, is likewise a discrete genotype.

Everything else is probabilistic. Polygenic embryo screening extends selection to traits shaped by thousands of variants, and the expected shift it can produce is a fraction of the variation within a single family. The constraint is not the accuracy of the laboratory work but the small genetic distance between siblings and the small number of embryos a cycle yields.

The arithmetic that popular coverage skips

A couple choosing among a handful of their own embryos is sampling from a narrow distribution. Even a perfect predictor of a trait could only pick the best of what those particular gametes happened to produce, which for most traits is a modest range.

§What editing would add, and what it would not

Editing changes the constraint. CRISPR–Cas9 and its derivatives can in principle install an allele neither parent carries, which selection can never do. That matters in a small number of situations where selection genuinely fails: when one parent is homozygous for a dominant disease allele, when both are homozygous for a recessive one, or when a couple produces too few embryos for selection to have anything to work with. These cases are real but rare, and they are the strongest medical argument for heritable editing.

For trait enhancement, editing runs into the same wall as selection, from a different direction. Traits such as height or cognitive test performance are influenced by thousands of variants, each shifting the outcome by a tiny amount, so a noticeable change means editing at a scale nobody has attempted in an embryo of any species. The number of changes required and the tolerance of a one-cell embryo for damage point in opposite directions. Base editing and Prime editing, which rewrite a base without severing both strands of the helix, are the chemistries that would have to carry such an attempt; both were developed for single, well-characterised disease alleles rather than for thousands of small-effect ones. Nuclease editing at several sites at once carries a documented risk of the large deletions and rearrangements that a biopsy of a few cells would not reliably detect.

Embryos add a further problem. Editing a one-cell zygote often produces a mosaic, an individual whose cells do not all carry the intended change, and mosaicism cannot be reliably detected by biopsying a few trophectoderm cells. Research on non-viable and research-consented human embryos has reported edits at target loci, though the mechanism claimed in the most publicised case was disputed by other groups who argued that apparent correction reflected allele dropout rather than repair.³ The He Jiankui affair of 2018 illustrated the whole set of failures at once: the intended CCR5 deletion was not faithfully reproduced, at least one child was mosaic, and the medical rationale did not hold up.

§Polygenicity as the real constraint

The gap between the popular image and the genetics is almost entirely explained by polygenicity. Public discussion imagines a menu of traits with switches attached, because Mendelian disease provides the mental model and Mendelian disease is genuinely switch-like. Nearly everything people would want to choose is not.

This is the central point of Genetic enhancement of cognition and applies equally to athletic ability, temperament and appearance. The variants involved are numerous, individually negligible, context-dependent, and pleiotropic, meaning the same variant influences several traits at once. A technology that can flip one switch cleanly is not on a continuum with a technology that can retune ten thousand dials.

What has actually been done

As of 2026 the only publicly documented births following deliberate genome editing remain the three children of the He Jiankui experiment. Everything else described as a designer baby has involved selecting among embryos, choosing a gamete donor, or, in the distinct case of Mitochondrial replacement therapy, substituting mitochondria rather than editing nuclear DNA.

§Ethics and law

The ethical literature separates arguments that apply to selection from those that apply to editing. Selection debates centre on the expressivist objection developed within Disability rights and enhancement, which holds that screening out a trait expresses a judgement about people who have it, and on Procreative beneficence, Julian Savulescu's principle that parents have reason to choose the child expected to have the best life.⁴ Editing debates add the questions of consent from a person who does not yet exist, of irreversibility across generations, and of the therapy-enhancement boundary examined in Human enhancement.

Bioconservative critics, surveyed in Bioconservatism, argue that treating a child's characteristics as a parental design choice alters the relationship between generations regardless of whether the technique works. Distributional critics make a different argument, set out in Access and inequality: a capability priced at the level of an IVF cycle plus genomic analysis is not evenly available, and one available only to the wealthy could compound advantage rather than health.

Law is fragmented. Heritable editing is prohibited outright in many countries and unfunded rather than banned in others; embryo testing is licensed condition-by-condition in some jurisdictions and unregulated in others. An international commission convened by national academies concluded that no clinical use of heritable editing should proceed until safety and efficacy could be established for a narrow set of indications, and that no country then had a pathway for authorising it.⁵ The details, and the weakness of enforcement, are covered in Governance of human genome editing.

§Outlook

Two technical developments would move the term closer to its popular meaning. In vitro gametogenesis, if it works in humans, would turn selection from a choice among a handful of embryos into a choice among hundreds, which is the only route by which polygenic selection could produce changes large enough to notice. Reliable multiplex editing of embryos, with mosaicism solved, would remove the parental-genome constraint entirely.

Neither is close, and the second may never be permitted. The open question is whether a capability that arrives gradually, one condition at a time and framed each time as disease prevention, ever confronts the public with the choice that "designer babies" names, or whether it simply passes without anyone having decided anything.

§See also

• Embryo selection
• Polygenic embryo screening
• Human germline editing
• He Jiankui affair
• In vitro gametogenesis
• Procreative beneficence
• Genetic enhancement of cognition
• Bioethics of enhancement

§References