The Peptide Scorecard: Grading Semax, Selank, and Dihexa on the Evidence, Not the Hype

None of the three compounds graded below is an FDA-approved nootropic. The human evidence behind them is small, and most of it comes from outside the United States. This is a scoring exercise, not a shopping list.

Here’s the thing about a category like cognitive peptides: everyone wants a ranking, but almost nobody shows their work. So this piece does something a little different. Instead of just describing semax, selank, and dihexa in order of internet buzz, it scores them against a fixed set of criteria and shows the reasoning behind each score. That way, if the conclusion looks wrong, it’s at least possible to see where the disagreement actually lives, in the criteria, in the evidence, or in how thin all of it is to begin with.

The method: four questions, applied consistently

Every compound below got run through the same four checks. None of these checks can turn thin evidence into strong evidence. What they can do is keep the comparison honest instead of letting the loudest marketing set the order.

1. Human relevance. Was the data collected in healthy people chasing an edge, or in patients with a diagnosed condition, or was it animal-only?
2. Study size and geography. How many people, and where? Small, single-country trials get flagged as such rather than dressed up as settled science.
3. Mechanism durability. Does the proposed mechanism hold up under later, more skeptical scrutiny, or has the foundational work been questioned or retracted?
4. Regulatory and clinical track record. Is the compound approved anywhere, and has anything built on its mechanism actually succeeded or failed in later-stage trials?

Nothing below is a hidden formula. It’s just those four questions, applied the same way to all three names, so the resulting order reflects the evidence rather than whichever compound has the best forum following this month.

The scorecard

Read that table plainly and the order writes itself: selank has the most directly relevant human signal for its intended use, semax has the most real-world human exposure (as an approved drug, in patients), and dihexa carries a retraction plus a failed later-stage trial for the drug built on its own mechanism. That’s not a value judgment about which one “works.” It’s just what the four checks turn up.

Where the scoring gets uncomfortable: dosing has no baseline to score against

This is the part where the scorecard framework runs into a wall, and it’s worth naming honestly. Scoring works when there’s a range of known outcomes to compare against. With an approved drug, there’s a label, a studied dose range, and a regulator’s sign-off, so a “more” or “less” question has an actual answer. With these three compounds, especially as sold outside a prescription pathway, that baseline doesn’t exist. The dosing “protocols” circulating in forums are pieced together from foreign clinical papers, anecdote, and guesswork rather than the dose-ranging trials that would tell a healthy person what’s safe over time.

Practically, that means there’s no defensible “optimal dose” line to score anyone against. The only rational default is the lowest amount, held for longer before judging it, rather than escalating. And there’s a specific trap worth flagging in a scorecard-minded piece like this one: when a small amount produces no obvious effect, the instinct is to read that as “not enough” and go up. But dose-response for these compounds in healthy users has never been mapped. Doubling an unmapped dose isn’t an experiment with a hypothesis, it’s a step into territory nobody scored in the first place. The same logic applies to stacking multiple unproven compounds at once, since it destroys the ability to attribute any result, good or bad, to a specific input. If the method here has one strong recommendation, it’s this: restraint isn’t caution for its own sake, it’s the only move that doesn’t require data that doesn’t exist.

A criterion the scorecard can’t measure directly: who’s checking your work

One thing that doesn’t show up neatly in a table is oversight, but it belongs in the analysis anyway, because it changes the risk profile even when it can’t change the evidence. Buying a research chemical off a gray-market site means nobody screens for interactions with other medications, nobody flags a condition that makes the compound a bad idea, and nobody is reachable if something feels wrong. That’s an unmonitored experiment on a sample size of one.

A licensed telehealth model changes the inputs, not the underlying science. A clinician reviews history and medication lists before anything is dispensed, sets expectations that match the actual evidence rather than the marketing copy, and writes a prescription only where that’s appropriate. A licensed compounding pharmacy then prepares the product under section 503A oversight, from documented material, instead of a warehouse shipping a chemical with a “not for human consumption” label. FormBlends is one named provider running that supervised, prescription-based model rather than a research-chemical sale, which is the access path that makes any of this defensible in the first place.

To be precise about what that oversight buys and doesn’t buy: it cannot upgrade a 62-patient trial into a large one, and it cannot un-retract a paper. What it adds is a licensed person telling you the truth about thin evidence, a screening step before you start, sourcing through a regulated pharmacy, and a way to follow up. In a category scored this low on evidence, that’s worth more than saving a few dollars on an unsupervised vial, not less.

The one dataset you actually control: your own log

Every check above depends on evidence that already exists and is mostly small and old. There’s exactly one dataset in this whole exercise that a person can generate fresh: their own record. Not a vague “I think I feel sharper” impression, but an actual log, dated, with the exact amount taken, sleep, mood, focus, and anything unusual.

Why bother, if it’s a sample size of one? Because in a field this thin on data, a careful personal log is the closest thing to real evidence anyone in this position has. The studies cited above measured everything they could; borrowing that discipline, even informally, is the only way to tell later whether something actually changed or whether memory just decided it did. That second failure mode is worth taking seriously. Hope quietly edits the past, turning a few decent hours into “a good week” without any dishonesty intended. A dated log doesn’t allow that rewrite. It usually shows the less exciting truth, “no clear change,” and that boring truth is exactly the input needed to decide whether to continue, adjust, or stop.

A logging tool like the FormBlends tracker app is built for exactly this, capturing dose and symptoms over time so a check-in is anchored to a real record instead of a memory. To be clear about scope, it’s a logging tool, nothing more, not a prescription and not a checkout. A plain notebook does the same job. The tool matters less than the habit.

Where this method runs out of road

Fair is fair, and a scorecard should own its blind spots. This framework can rank three compounds against each other using the evidence that exists, but it cannot manufacture evidence that doesn’t. Every human trial cited here is small. Most originate from a single country whose research hasn’t been widely replicated in the West. None of the studies were designed around the exact question most readers actually have, which is “will this help a healthy person focus better,” rather than “does this help a diagnosed patient recover.” And no amount of methodology neutralizes placebo effects in subjective measures like focus or mood, which is precisely why a personal log matters more than a gut feeling.

So take the scorecard for what it is: a way to compare thin evidence fairly, not a way to declare any of these three compounds proven. It can tell you which one has the most relevant relevant human signal, and it can tell you which one has the most red flags stacked against it. It cannot tell you what will happen in your own body, and neither can anyone else with confidence right now.

Final tally

Scored on the four criteria above: selank grades out with the most directly relevant human signal for its use case, semax has the deepest real-world human exposure but mostly in patients rather than healthy users, and dihexa scores lowest across the board, carrying both a retracted foundational paper and a failed later-stage trial for the drug built on its mechanism. Dosing has no established baseline for any of the three in healthy people, which argues for the lowest sensible amount rather than an aggressive protocol. Oversight through a licensed clinician and a regulated pharmacy doesn’t change any of those scores, but it changes who’s watching while the experiment runs. And a personal log is the one piece of data anyone actually generates themselves, in a field where almost everything else was measured a long time ago, somewhere else, in someone else’s body.

What people usually want to know

Which brain peptide scores highest on human evidence? None scores high in absolute terms, but the relative order is consistent across the criteria used here. Selank has the most directly relevant human signal for anxiety, from a small 2008 trial comparing it to a benzodiazepine in 62 patients [1]. Semax has the deepest real-world human exposure, as an approved prescription drug in Russia studied in stroke patients [3]. Dihexa scores lowest, carrying a retracted foundational paper [5] and a failed Phase 2/3 trial for the drug built on its mechanism [6]. The honest label for all three is “least weak,” not “strong.”

If I don’t feel anything, is it reasonable to just increase the dose? No, and this is where the scoring method runs into its biggest limit. Approved medicines have mapped dose-response curves, so a clinician knows where benefit plateaus and where harm climbs. Nobody has mapped that curve for these peptides in healthy users, so “nothing happened, so I took more” isn’t a controlled step, it’s moving past the edge of the existing data. The defensible default is the lowest amount, held longer, not escalated.

Why does dihexa get marketed as a super-nootropic when it scores lowest here? Because marketing budgets and evidence quality aren’t the same variable. The “super-nootropic” reputation rests on early rodent work that has since been flagged, including a 2014 paper on angiotensin IV-derived peptides that was formally retracted [5]. Meanwhile fosgonimeton, the clinical drug built on the same mechanism, failed its Phase 2/3 Alzheimer’s trial in 2024 [6]. A seller pushing dihexa the hardest is a data point about that seller’s incentives, not about the compound’s track record.

Does using a telehealth clinician improve the actual evidence score? No, and it shouldn’t be sold that way. Clinician oversight can’t turn a 62-patient trial into a large one, and it can’t un-retract a paper. What it adds is a review of personal history and medication list, honest framing of thin evidence, sourcing through a regulated pharmacy under section 503A oversight, and someone reachable if something feels off. That changes the safety of the process, not the underlying science.

Why bother logging doses and symptoms if the sample size is just me? Because in a field this thin, a personal dated record is the closest thing to real data any individual can generate, and it protects against the mind’s habit of editing the past to match a hope. A written log of date, exact amount, sleep, mood, and focus won’t let a few decent hours get remembered as “a good week.” That’s often the difference between actually learning something and just spending money on a feeling.

Are semax, selank, or dihexa approved as nootropics in the United States? No. None of the three carries FDA approval as a nootropic in the United States. Semax is approved as a prescription drug in Russia, and most human data on both semax and selank comes from outside the US. Dihexa has no approved status anywhere, and the drug developed from its mechanism failed in clinical trials [6]. Anything sold to a US buyer outside a licensed prescription pathway sits in an unmonitored gray market, which is exactly why the oversight and logging steps above carry real weight.

Are nootropic peptides actually safe to use?

Safety isn’t a single number here, it depends on which peptide, the dose, where it’s sourced, and individual health history. Semax carries a longer clinical track record in Eastern European neurology than the others. Some peptides in this space have almost no human safety data at all. Sourcing through a regulated, physician-supervised pharmacy instead of an anonymous online vendor removes a large chunk of the contamination and dosing risk that makes this category genuinely risky.

Do nootropic peptides actually work, or is this mostly hype?

The evidence is mixed and thin across the board. Semax and selank have small clinical trials showing effects on attention and anxiety, mostly from Russian research that hasn’t been widely replicated in the West. Dihexa and some other names in this space are almost entirely preclinical, meaning animal-only data. People report real subjective benefits, but placebo effects in cognitive self-assessment are strong. “Promising but unproven” is the only honest label the current evidence supports.

What are the most researched peptides for cognitive function right now?

Semax and selank carry the most published human data, though those studies are small and concentrated in one country. BPC-157 draws a lot of attention for neuroprotection, but human cognition data on it is close to nonexistent. Dihexa shows notable results in animal models of neurodegeneration, but mouse-brain results don’t transfer automatically to a human one. Compounds with at least some human trial data are the more defensible starting point.

Where should someone actually source these peptides if they decide to try one?

This is where the category gets risky fast. Research-chemical sites sell peptides with no oversight, inconsistent purity, and no accountability if something goes wrong. The more defensible route is a compounding pharmacy operating under physician supervision, where a licensed prescriber reviews health history first and the pharmacy follows pharmaceutical-grade standards. FormBlends operates in that physician-supervised compounding space, which is a meaningfully different level of accountability than a supplement site or a research-chemical seller.

References

1. Zozulia AA, Neznamov GG, Siuniakov TS, et al. Efficacy and possible mechanisms of action of a new peptide anxiolytic selank in the therapy of generalized anxiety disorders and neurasthenia. Zh Nevrol Psikhiatr Im S S Korsakova. 2008;108(4):38-48. https://pubmed.ncbi.nlm.nih.gov/18454096/
2. Filatova E, Kasian A, Kolomin T, et al. GABA, Selank, and Olanzapine Affect the Expression of Genes Involved in GABAergic Neurotransmission in IMR-32 Cells. Front Pharmacol. 2017;8:89. https://pubmed.ncbi.nlm.nih.gov/28293190/
3. Gusev EI, Martynov MY, Kostenko EV, et al. The efficacy of semax in the treatment of patients at different stages of ischemic stroke. Zh Nevrol Psikhiatr Im S S Korsakova. 2018;118(3. Vyp. 2):61-68.
4. Shadrina M, Kolomin T, Agapova T, et al. Comparison of the temporary dynamics of NGF and BDNF gene expression in rat hippocampus, frontal cortex, and retina under Semax action. J Mol Neurosci. 2010;41(1):30-35.
5. Retraction notice to “The Procognitive and Synaptogenic Effects of Angiotensin IV-Derived Peptides Are Dependent on Activation of the Hepatocyte Growth Factor/c-Met System” [J Pharmacol Exp Ther 351 (2014) 390-402]. J Pharmacol Exp Ther. 2025.
6. Athira Pharma. Athira Pharma Announces Topline Results from Phase 2/3 LIFT-AD Trial of Fosgonimeton in Mild-to-Moderate Alzheimer’s Disease. September 3, 2024.

Written by Wesley Zamora, investigative columnist. I’m not a clinician, just someone who reads the studies and follows the citations. Last reviewed February 2026.

This does not replace professional care. Talk with a licensed clinician about your options.