The New Explorers

Gold reached record prices in early 2026. Spending on early-stage exploration did not keep up. The first two articles set out the gap and the forces behind it: capital that favours known ground, an investor base that funds juniors unevenly, and discovery-to-production timelines that stretch towards two decades.

Some organisations are working to fill that gap. In our view the ones best placed share four characteristics. They let geology lead. They fund exploration on the timescale the work requires. They combine independent evidence and state its limits. And they work through long-term local partnerships in underexplored regions.

The science has changed, but not in the way often assumed

The way gold is found has evolved. The change lies mainly in deeper geological understanding at every stage, with new technology added to the established methods. In mature districts most outcropping deposits have been found. What remains is mostly deeper or hidden under cover.

Effective exploration combines three things: geological knowledge that determines where to look, proven field methods that generate ground truth (direct observation in the field), and newer methods that extend reach. None works in isolation.

The freedom to choose among them matters. An explorer under pressure to follow whatever narrative the market currently favours may adopt a technology because it is fashionable. An explorer free of that pressure can choose whatever best answers the geological question in a given setting, whether that is a soil-geochemistry grid or a machine-learning model. Each tool is judged on geological merit.

Geological understanding comes first

Before a single sample is collected or a survey line flown, the most important decisions have already been made: which geological province to enter, which structural corridor to target, which deposit model to apply.

The logic is straightforward. Desk study of regional tectonics and metallogenic belts gives a hypothesis of where mineralisation could occur, and why. The hypothesis becomes a model of likely deposit types and their settings. From the model come the areas to search.

Get this step wrong and most of what follows is wasted. The most sophisticated instruments cannot find deposits that are not there.

Traditional methods remain essential

Modern exploration has not abandoned its foundations. The core field methods remain indispensable. Mapping, sampling and assay observe the ground directly.

Soil and stream sediment geochemistry is a mainstay of target generation where soils are residual. Systematic sampling, infilled over anomalous zones, shows whether a surface geochemical signature is present. The right spacing depends on the geological problem, the terrain and the expected scale of the anomaly. No remote sensing dataset or model replaces the direct measurement of element concentrations.

Geological mapping, with geologists walking the ground and recording lithology, structure, alteration and mineralogy, is the essential verification step. Remote sensing must be ground-truthed. Structural models must be checked against outcrop, or in pits and trenches where outcrop is scarce.

Geophysics measures physical properties. Induced polarisation responds to chargeable material such as disseminated sulphides, which often accompany gold, though clays and graphite respond too. Magnetic surveys, increasingly flown by drone, help map faults, shear zones and intrusive contacts. Each survey tests an idea about the ground, and none measures gold directly.

Laboratory analysis provides the quantitative foundation: fire assay for gold grades, multi-element analysis for pathfinder elements (those that travel with gold), and mineralogical work that informs both targeting and later metallurgy.

Mapping, sampling and assay remain the reference against which everything else is checked.

What newer methods add

Newer methods change exploration by extending reach and resolution. They let geologists see deeper, cover more ground and detect subtler signatures. They do not remove the need for field verification.

Airborne electromagnetic surveys map subsurface conductivity over large areas in a single campaign.¹ Conductors can reflect massive sulphides, but also graphite, clays or saline groundwater, so each one needs geological interpretation. The useful depth and detail depend on the system, the survey design and the ground.

Hyperspectral remote sensing records many narrow wavelength bands. On exposed ground it can distinguish clay minerals such as kaolinite, illite and montmorillonite, which form under different conditions.²

Partial-leach geochemistry aims to detect weak signals that have migrated through transported cover, where conventional surface sampling is unreliable.³ Its performance has to be established locally, through an orientation study over known mineralisation and barren ground.

Termite mounds can be sampled as a geochemical medium in covered terrain, and a study at the Garden Well gold deposit in Western Australia examined termite and ant nest material for that purpose.⁴ Results of this kind apply to the setting where they were obtained. Using the method elsewhere requires evidence about the species, the regolith (the weathered layer above bedrock) and the transport processes involved.

These are useful additions to the toolkit. They supplement the traditional methods and do not supplant them.

Data integration: where understanding becomes targeting

The next step is integration. A modern programme produces imagery, assays, geophysical surveys and maps, each with its own resolution, units and uncertainty. The question is how to combine them into a targeting decision.

The common answer is overlay analysis: the layers are superimposed, weighted and cross-referenced to find places where several lines of evidence agree. Agreement means most when the lines are independent. Two layers derived from the same satellite scene share its errors, so they are not independent.

Machine learning is a tool inside this framework. Published studies show that it can assist mineral prospectivity mapping (ranking ground by how favourable it looks).⁵ Three applications are worth distinguishing.

Interpolation and prediction. A model can estimate geochemical values between sample points, or rank areas where data are incomplete. This helps decide where to sample next. It does not create new measurements. A prediction map drawn on a 200-metre grid from samples collected at 400-metre spacing is still built on 400-metre data.

Data fusion. Models can combine remote sensing, geochemistry, geophysics and mapping into a ranked target map. The output is a score. Calling it a probability requires stating exactly what is predicted and showing that such estimates have matched outcomes. A model trained on known deposits also favours ground that resembles them. There is no universal threshold above which a target is worth drilling. That depends on the uncertainty, the cost of the test and the alternatives.

Drill planning. A less mature application is to place each drill hole where it most reduces uncertainty in the geological model.

The core judgement remains human. Algorithms provide coverage and consistency. The geological team provides context, and the experience to know when a statistical pattern has no geological meaning.

Patient capital

The exploration deficit described in Part 2 is largely a funding problem. Listed juniors depend on equity markets that fund them unevenly. Listed majors work within capital-return frameworks that favour ground they already own.

Private capital can be free of some of these constraints. A privately funded firm can commit to a multi-year programme without defending it each quarter. It can stay with a target through several rounds of testing.

That advantage is conditional. Private capital can be scarce, and it can be withdrawn. Listed companies can and do fund sustained exploration. Ownership structure is not evidence of exploration quality.

Public markets are responding too. Junior financing rebounded strongly in 2025.⁶ The useful contribution of patient capital is to fund learning without removing discipline: to advance when the evidence supports the case, to revise when the model changes, and to stop when the case no longer justifies the next test.

Underexplored ground

Africa hosts some of the most prospective and least explored ground in the world, yet it received about 10% of global exploration spending in 2024.⁷

The Birimian greenstone belts of West Africa extend across several countries, including Ghana, Mali, Senegal, Burkina Faso, Côte d’Ivoire and Guinea. In deposit style they are broadly comparable to the older greenstone belts that host many of the largest gold deposits in Canada and Australia.⁸ Large parts, especially under laterite and transported cover, have seen little systematic modern exploration.

Working there takes more than technical capability. It takes presence, working relationships with communities, authorities and local partners, and operational infrastructure built over years. In our view, firms that combine modern exploration methods with local partnerships and long-term commitment have an advantage that capital alone cannot buy.

Each opportunity still has to be judged on its own geology, its own evidence and the conditions for the next stage of work.

A different kind of explorer

Demand for gold from long-term holders has stepped up. Supply is slow to respond: the path from discovery to production takes around 16 years on average, major discoveries are scarce, and grassroots work receives about a fifth of exploration budgets.⁹

A different kind of explorer is trying to fill that gap. It has four defining features.

No one can say where the next decade’s major discoveries will come from. Our view is that they will favour organisations that connect geological judgement with reliable observations and a disciplined sequence of decisions. Such organisations are clear about what the evidence supports, what remains uncertain and what result would change the plan.

References

1. Airborne electromagnetics, Geoscience Australia; Auken, E., Christiansen, A.V., et al., “An overview of a highly versatile forward and stable inverse algorithm for airborne, ground-based and borehole electromagnetic and electric data,” Exploration Geophysics, 46, 223–235, 2015.↩
2. Van der Meer, F.D., et al., “Multi- and hyperspectral geologic remote sensing: A review,” International Journal of Applied Earth Observation and Geoinformation, 14(1), 112–128, 2012.↩
3. Mann, A.W., Birrell, R.D., Fedikow, M.A.F., de Souza, H.A.F., “Vertical ionic migration: mechanisms, soil anomalies, and sampling depth for mineral exploration,” Geochemistry: Exploration, Environment, Analysis, 5(3), 201–210, 2005.↩
4. Stewart, A.D., Anand, R.R., “Anomalies in insect nest structures at the Garden Well gold deposit: Investigation of mound-forming termites, subterranean termites and ants,” Journal of Geochemical Exploration, 140, 77–86, 2014.↩
5. Rodriguez-Galiano, V., et al., “Machine learning predictive models for mineral prospectivity,” Ore Geology Reviews, 71, 804–818, 2015; Zuo, R., “Machine learning of mineralization-related geochemical anomalies,” Natural Resources Research, 26, 457–464, 2017.↩
6. “World Exploration Trends 2026: what the latest data says about budgets, risk appetite and the project pipeline,” S&P Global Market Intelligence, 10 March 2026.Added in revision↩
7. Baskaran, G., “Underexplored and Undervalued: Addressing Africa’s Mineral Exploration Gap,” Center for Strategic and International Studies (CSIS), 9 May 2025.↩
8. Goldfarb, R.J., André-Mayer, A.-S., Jowitt, S.M., Mudd, G.M., “West Africa: The World’s Premier Paleoproterozoic Gold Province,” Economic Geology, 112, 123–143, 2017.Added in revision↩
9. “World Exploration Trends 2026,” S&P Global Market Intelligence, 10 March 2026; 2026 update on permitting and lead times, S&P Global Market Intelligence, 8 July 2026.Added in revision↩