If you want the short answer first, here it is: most mainstream guidance puts gold at 5-15% of a portfolio, and 10% is the single most cited figure. The honest longer answer is that the right percentage depends on the job you hire gold to do, which is why credible recommendations run all the way from 2% to 20%.
What follows is an educational framework, not a prescription: what the research supports, why institutions hold far less than the studies suggest, how 2025 and 2026 changed the debate, and how to keep your chosen number from drifting. Set the final allocation with your own adviser, against your own balance sheet.
The academic and institutional work on gold allocation is more consistent than most investors expect. The World Gold Council found that allocations between 4% and 15% improved risk-adjusted returns for typical multi-asset portfolios over long horizons. WisdomTree's portfolio optimization research put the "optimal" weight near 13%. At the aggressive end, CPM Group has argued for 20%, though that figure deserves a flag: CPM's work is sponsored by the gold industry, so treat it as the ceiling of advocacy rather than the center of evidence.
Now the other side of the ledger. UBS surveys of family offices put their average gold and precious metals exposure at roughly 2%. JPMorgan estimates that investors overall hold about 2.8% in gold. In other words, the wealthiest and most professionally advised portfolios in the world hold a fraction of what the studies say is efficient.
The studies cluster between 4% and 15%. Actual investors hold closer to 3%. The gap between what the research supports and what people actually own is the real story of gold allocation.
Part of that gap is opportunity cost. Per JPMorgan's own long-run numbers, the S&P 500 returned roughly 11.9% per year from 1985 to 2025 against about 6.7% for gold. But over 2005 to 2025 the order flips: gold at roughly 11.6% per year versus about 10.7% for stocks. Gold wins the last 20 years; stocks win the last 40. Which window you anchor on quietly decides how much gold you think you need.
Rather than arguing about the "correct" percentage, it is more useful to decide what job gold has in your portfolio. Each job implies a different size.
Here gold is a hedge against the scenarios you hope never happen: currency crises, geopolitical shocks, deep equity drawdowns where correlations spike. An insurance sleeve is deliberately small because you expect it to underperform most of the time. If this is the job, 2-5% is coherent, and complaining that it "lags stocks" misses the point. Insurance is supposed to cost something in good years.
This is the mainstream case and the range where most of the research sits. Gold's long-run correlation to stocks and bonds is low, and in years when both fall together, as they did through 2022, an uncorrelated asset earns its keep. A 5-10% sleeve is large enough to move portfolio-level volatility and drawdowns, small enough that a bad decade for gold does not sink the plan.
Some investors hold a structural macro view: persistent central bank buying, fiscal deficits, de-dollarization, a weaker case for bonds as the ballast asset. If that is your genuine thesis, 10-15% or more is internally consistent. But be honest that this is an active bet, not diversification, and size it like one. This is also the zone where the industry-sponsored research tends to live.
For decades, "5 to 10 percent in gold" was a slightly contrarian statement. Three things pushed it into the mainstream.
Central banks became the marginal buyer. Central banks bought 863 tonnes of gold in 2025, per the World Gold Council, with Poland alone adding 102 tonnes. The WGC's 2026 survey found a record 45% of central banks planning to increase gold reserves, and the council forecasts roughly 850 tonnes of official buying again in 2026. That structural bid is a large part of why gold returned about 65% in 2025, its best year since 1979, on top of 27% in 2024. For a deeper look at what actually moves the metal, see what drives gold prices.
The stock-bond correlation regime broke, then healed only partly. From early 2022, investors lived through 14 consecutive months in which stocks and bonds fell together, and the 12-month stock-bond correlation reached 0.80 by July 2024 before easing to about 0.16 by September 2025, per State Street. When bonds stop reliably hedging equities, the case for a third diversifier strengthens. That debate is the subject of our piece on whether the 60/40 portfolio is dead, including the "60/20/20 with gold" construction that strategists floated on CNBC in October 2025 and Morgan Stanley's 20% gold call.
Then gold reminded everyone it is volatile. The metal hit an intraday record of $5,589 per ounce on January 28, 2026, and has since corrected roughly 25%, trading near $4,121-$4,143 in the week of July 6-10, 2026. It is still up about 23% year over year, but anyone who chased the top has already lived through a drawdown deeper than most equity corrections. Where it goes from here is a separate question; our gold price forecast for 2026-2027 tracks the major bank targets. The allocation question does not depend on the forecast being right.
Once you have a number, there are three broad implementation routes, and they are not interchangeable.
The least discussed part of gold allocation is that the number you pick will not stay picked. A 10% sleeve entering 2025 would have swollen well past target after a 65% year, exactly in time for the 25% correction of 2026.
The arithmetic of drawdowns is unforgiving and worth memorizing: a 10% loss needs an 11.1% gain to recover, a 20% loss needs 25%, a 33% loss needs 50%. A sleeve that drifts from 10% to 16% of the portfolio does not just carry more upside; it carries 60% more exposure to that math.
A simple discipline is a relative band: rebalance whenever the sleeve drifts more than about a quarter away from target, so a 10% target gets trimmed above 12.5% and topped up below 7.5%. Vanguard's Advisor's Alpha research attributes a modest but real annual benefit, on the order of tenths of a percent, to disciplined rebalancing. The bigger benefit is behavioral: the band forces you to sell some gold into euphoria and buy some into fear, which is precisely what most investors fail to do on their own.
Gold is not for every portfolio, and an honest framework says so. You probably belong at the low end, or at zero, if:
The framework, then: pick the job, pick the matching range, implement it in the cheapest structure you will actually maintain, and set bands so the sleeve stays the size you chose. Whatever number comes out of that process, pressure-test it with a qualified adviser who can see your whole financial picture. The worst gold allocation is the one chosen at a price peak and abandoned in a correction.
Not according to most of the research. World Gold Council analysis found allocations of 4-15% improved risk-adjusted returns, and 10% sits comfortably inside that band. It is, however, well above what most investors actually hold, roughly 2.8% per JPMorgan estimates, so 10% is a deliberate diversification decision rather than a default. Whether it is right for you depends on your horizon, income needs, and what else is in the portfolio.
Buffett is famously skeptical of gold. His core argument is that gold is a non-productive asset: it generates no earnings, dividends, or interest, so its return depends entirely on someone paying more for it later, while businesses and farmland compound value internally. That critique is intellectually serious and worth understanding even if you hold gold. The counterargument is that gold is not owned for compounding; it is owned for diversification and crisis behavior, a job productive assets do not always perform.
For most portfolios under a few million dollars, ETFs are the practical choice: lower all-in costs, instant liquidity, and easy rebalancing. Physical gold removes counterparty and financial-system risk but adds dealer spreads, storage, insurance, and friction. Many larger allocators split the difference, holding a small physical core for insurance and using ETFs for the rebalanced portion of the sleeve. Tax treatment differs by structure and jurisdiction, so check that with a professional before choosing.
A common approach is a relative band of about 25% around target: for a 10% allocation, trim above roughly 12.5% and add below roughly 7.5%, checked quarterly or after large moves. The 2025-2026 cycle shows why: a 65% up year followed by a roughly 25% correction would have carried an unrebalanced sleeve far above target and then through an outsized drawdown. Calendar-only rebalancing works too; the key is having a rule you follow in both directions.
They do different jobs. Bonds provide income and typically cushion recessions; gold provides no income but has historically helped in inflation shocks and periods when stocks and bonds fall together, as in 2022 when the stock-bond correlation regime broke down. State Street data shows that correlation reached 0.80 in mid-2024 before normalizing, which is exactly when a third diversifier proves useful. Most research treats gold as a complement to bonds, not a replacement, which is the logic behind the 60/20/20 constructions discussed in 2025.