Is intelligence heritable? Yes. Most public discussion gets the science wrong.
If you're going through IVF and are testing your embryos for the risk of developing complex diseases, you may wonder whether it’s also possible to test for intelligence. The answer is yes: intelligence is moderately to highly heritable (Plomin & Deary, 2015), and Herasight has developed a powerful polygenic predictor of intelligence (Wolfram et al., 2026).
Cognitive ability shapes your child's life more than most parents realize
Intelligence not only correlates positively with income and educational attainment (Strenze, 2007). It is also associated with health (Fries & Pietschnig et al., 2025) and lifespan (Sánchez-Izquierdo et al., 2023), and seems to reduce the risk of addictive behaviors such as smoking (Batty et al., 2007) and alcohol abuse (Capusan et al., 2025) . This makes sense because intelligence is correlated with self-control and long-term planning (Shamosh & Gray, 2008).
The genetic component is the one that doesn't fade out
Heritability estimates for cognitive ability in adults can be as high as 80% (Plomin & Deary, 2015). Estimates are lower in childhood and higher in adulthood. Exactly why heritability increases with age still isn't fully understood, although the pattern itself is highly replicated.
Large-scale genome-wide association studies like Savage et al. (2018), which analyzed nearly 270,000 individuals, have identified hundreds of genetic variants associated with intelligence. Those variants are not randomly scattered across the genome. They're enriched in conserved and coding regions, and the associated genes are strongly expressed in the brain, specifically in neurons involved in learning and memory.
The genetic aspect of intelligence is real and can be used for selection. One benefit of embryo screening for intelligence is that the benefits will be meaningful and long lasting, in contrast with environmental interventions. A meta-analysis of 141 educational RCTs involving more than one million students, Lortie-Forgues & Inglis, 2019, found an average effect size of just 0.06 standard deviations for improvement in educational outcomes. Morover, another issue is that gains from environmental interventions such as Head Start fade within a few years (Protzko, 2015), resulting in a loss of the gains. Even schooling, which appears to add a few IQ points per year of enrollment (Ritchie & Tucker-Drob, 2018), most likely produces skill-specific rather than general gains (Ritchie, Bates & Deary, 2015). The test scores increase, but general ability does not.
We can now measure it between sibling embryos
Five years ago, the best publicly available cognitive predictors would have yielded roughly 2.5 IQ points of expected gain from screening 10 embryos (Karavani et al., 2019). The authors called that "limited utility." They weren't wrong. The scores captured only a small fraction of the relevant genetic variation, but a lot has changed since then.
The improvement came from the entire pipeline getting better: larger training datasets, smarter phenotyping (the way you define and measure cognitive ability matters more than most people realize), new meta-analytic methods, and better statistical models. One piece of our approach is SBayesRC (Zheng et al., 2024), which uses information about what different parts of the genome actually do rather than treating every variant identically. The metric that matters most for embryo screening is within-family prediction. Think of your embryos as potential siblings. Siblings share parents, ancestry, and (eventually) the same household. Population-level studies pick up signal from all the things that differ between unrelated people: culture, neighborhood quality, parental education. That noise inflates apparent accuracy, especially for traits like cognition which can be influenced by social factors and the home environment. Population confounding is exactly why within-family validation matters more for cognition than for almost any other trait. If a score still predicts which sibling scores higher after stripping everything they share, the remaining signal is genuinely genetic.
CogPGT showed a within-family correlation of 0.456 with general cognitive ability in UK Biobank adults (Wolfram et al., 2026). Full methods are in our CogPGT technical paper and the preprint.
Five years ago, it was more reasonable to call the gains "limited utility." But this is clearly no longer the case.
We have additional evidence to suggest that the predictor is capturing real genetic signal, rather than capturing socioeconomic confounding: In the ABCD cohort, we found no significant interaction between CogPGT and family income, parental education, or family conflict.
Concerns about a relationship between intelligence and autism are common. Older population-level work reports positive genetic correlations between cognitive measures and autism (Grove et al., 2019; Hagenaars et al., 2026). The best current hypothesis is that intelligence is mainly correlated with the systematising component of autism (the Asperger's end of the spectrum) rather than with low-functioning autism. In the ABCD data, CogPGT showed a slightly negative association with autistic symptoms. Higher cognitive scores were not linked to more autism risk. But the honest answer is that we don't fully understand the relationship between intelligence and autism yet.
What IVF families should do with this
If you have only one euploid embryo and do not want to do additional cycles, screening for intelligence changes nothing. However, if you have enough euploid embryos that there's a real choice, cognitive screening is rational to include if cognitive outcomes matter to you.
In practice, polygenic embryo screening becomes most useful when you have roughly twice as many euploid embryos as desired children. Fewer can still matter. More is better. The point is having substantial variation to work with.
Nobody can create traits outside the parents' genetic range. But embryos can differ substantially in the set of disease risks they face, as well as their cognitive potential. Screening makes some of those inherited differences visible before transfer.
Intelligence is heritable, and can be predicted for the same reasons traits like height and disease risks like diabetes can be predicted. But not every company can predict these traits equally, and Herasight has developed by far the most powerful predictor of cognitive ability in the world (Wolfram et al, 2026). Intelligence affects your child's health, education, and lifespan in ways that are causal, not just correlational. And unlike environmental interventions, the genetic component of intelligence doesn't fade out. When you have several healthy embryos to choose between, using that information can be valuable.
If you're curious what cognitive screening could look like for your specific embryo count, the expected gains calculator lets you see the numbers. If you want to talk through what any of it means, reach out to our team.