Gold Mining Technology: 10 Methods With Real Data

Gold mining technology cuts the cost and risk of moving and processing rock, but it does not set the gold price. The World Gold Council (WGC) put the global average all-in sustaining cost (AISC, the yearly cost of keeping a mine running, including sustaining capital, royalties and closure costs) at US$1,785 per ounce in the first quarter of 2026, up 16% on a year earlier and the 28th year-on-year rise in a row (WGC, August 2026). Ten methods are compared below, each with the dated public data we found.

Ten gold mining technologies compared

The table gives the job each method does, the evidence we found and where it is strongest. Where a method has no public gold-mine data, the table says what it is for rather than inventing a saving.

How we chose these methods

The list keeps the ten methods of the original guide. A method with a company filing, a supplier milestone or a test report gets figures; one that only has general descriptions is described, not quantified. Every figure carries its source and date. Saving claims from equipment suppliers are labelled as supplier figures. Nothing here is a forecast, and costs differ by ore body, country and fleet size.

1. Autonomous haul trucks

Best for: large open pits with long, steady haul cycles. Komatsu says it commissioned its 1,000th autonomous ultra-class truck in April 2026, a 290-tonne payload 930E-5AT at Nevada Gold Mines, and that its customers have moved over 11.5 billion tonnes with the system (International Mining, April 2026). Nevada Gold Mines, run by Barrick as a joint venture with Newmont, announced in July 2025 that it would automate its 300-tonne and 230-tonne trucks, and Barrick’s then chief executive Mark Bristow said removing the onboard operator lowers exposure to hazards (Barrick and Komatsu release, July 2025).

Chart

Komatsu autonomous trucks by model

Autonomous trucks commissioned worldwide up to April 2026, approximate (from Komatsu’s 52% / 24% / 24% shares), all commodities

On productivity, the best-known figure is a supplier one. Caterpillar and Fortescue reported a 30% productivity improvement from autonomous haulage at the Fortescue iron ore operation in 2019 (Canadian Mining Journal, November 2019). That is iron ore, not gold, so treat it as an upper marker rather than a gold-mine promise. Freeport-McMoRan reported that Bagdad, its Arizona copper mine, finished converting to an autonomous fleet in 2025 (Freeport 10-K, 2025). Our look at automated mining and mine safety covers the wider picture.

Cons: the fleet needs a private network, mapped haul roads and a control room, so small pits may not suit it.

2. Sensor-based ore sorting

Best for: low-grade ore where waste rock can be rejected before grinding. Sorters scan lumps with X-ray, optical or other sensors and blow waste out with air jets. A 2019 AusIMM paper on AngloGold Ashanti ores reported gold recovery above 95% at a 40% mass pull (the share of feed sent on to the mill) using one or two sorting steps, and stressed that texture and the way gold sits in the rock decide the result (AusIMM, World Gold 2019).

Chart

Ore sorting by rock size, Cariboo test

X-ray sorting of about 80 tonnes of underground material

A newer, larger test comes from Osisko Development. An X-ray transmission sorter run on about 80 tonnes of Cariboo underground material rejected 42% to 55% of the rock by weight as waste in the 10-30 mm and 30-70 mm fractions while keeping 84% to 89% of the gold. For the 6-10 mm fraction, which AI advances made testable, it rejected 59% and kept 92% on an 8-tonne subsample (Osisko Development release, 7 July 2025, corrected). The chart shows the trade-off: finer rock sorts better, coarser rock gives up more gold.

Cons: gold lost in rejected rock never comes back, and a test on one ore body does not carry over to another. The calculator below shows when the saved milling cost outweighs the gold given up.

3. AI and machine learning

Best for: teams with years of clean sensor, assay and maintenance data. Machine learning looks for patterns in drill logs, plant readings and equipment data. Freeport’s filing says the company may apply AI to decision support, material characterisation, equipment reliability and extraction, which shows how operators describe it: a set of tools, not one product. The sorting tests above are a concrete gold example, as AI tools are what made the 6-10 mm fraction measurable. Our guide to automated drilling systems shows where the same tools reach the drill rig.

Cons: few operators publish before-and-after numbers, and a model trained on one deposit can mislead on another. Ask any vendor for results on ore like yours.

4. Drones and satellites for surveying and exploration

Best for: mapping terrain, stockpiles and surface clues before spending on drilling. The WGC lists drone surveys of sites among the ways a gold miner can cut transport overheads (WGC, Gold and climate change report). Satellites do a different job: they cannot see gold itself, but they map rust-coloured iron stains, clay minerals and fault lines that gold systems can leave behind. Both tools show the surface only, so drilling still decides what lies below.

Farmonaut screens a licence polygon from satellite images before ground crews go in. For the clues themselves, see our guide to finding gold deposits and the explainer on epithermal gold systems.

Cons: cloud, vegetation and soil cover can hide the signal, and a map is a target to check, not a resource estimate.

▶ Explains how satellites map rust stains, clay minerals and faults around gold systems, since they cannot see gold itself. · 4:25

5. Sensor networks and the Internet of Things

Best for: mines with expensive fixed plant and ventilation. Networked sensors on pumps, crushers, vehicles and air shafts feed one dashboard. The WGC notes that Internet of Things devices combined with predictive analytics can cut energy use, for example by reducing the need for back-up diesel generation (WGC, Gold and climate change report). Newmont also says it is building an energy management system, with a pilot planned for 2026, that is expected to cut energy use by 2% to 5% through asset efficiency (Newmont sustainability report, 2025).

Cons: sensors create data, not savings. Value comes only if someone acts on the alerts.

6. Processing upgrades: recovery and chemicals

Best for: refractory ore, where gold is locked inside sulphide minerals, and sites under pressure to cut chemical use. Gold is usually crushed, dissolved, caught on carbon, melted and refined. Pre-treatment steps such as bio-oxidation use bacteria to break down sulphide minerals so the gold can be reached, but the leach that follows normally still uses cyanide, so these methods reduce difficulty rather than replace the chemical. The video below follows one gold nugget through every stage.

Cons: the right circuit depends on the ore. Metallurgical test work on your own sample must come first.

▶ Follows one gold nugget from rock through drilling, crushing, leaching, carbon, melting and refining. · 5:27

7. 3D geological modelling

Best for: turning drill results into a block model (a 3D grid of cells, each holding an estimated grade) that supports a resource estimate and a pit or stope design. A model joins drill assays, structure and rock type into one volume, so planners can test cut-off grades (the lowest grade worth mining) before building anything. Software choices are compared in our guide to 3D geological modelling tools.

Cons: a model cannot be more certain than its drilling. Wide drill spacing gives wide error bands.

8. Renewable power and battery-electric trucks

Best for: remote sites that burn diesel. The WGC estimated that, for gold miners, purchased electricity made up roughly 40% to 45% of greenhouse gas emissions and diesel and fuel burned on site roughly 30% to 35% (WGC, Gold and climate change report, published in October 2019). Newmont reports that renewables were 22% of its purchased grid electricity in 2025, up from 21% in 2024, that Boddington has signed a wind power purchase agreement starting in 2026, and that its Brucejack mine has run a fleet of 50-tonne battery-electric haul trucks for several years (Newmont sustainability report, 2025).

Cons: power supply deals are site-specific, and battery trucks suit underground ramps better than deep open pits today.

▶ Walks through the seven environmental impacts of mining in plain words. · 2:31

9. Robotics, wearables and augmented reality

Best for: keeping people out of the most dangerous work. Driverless trucks already remove the operator from the cab, and remote-controlled loaders, charging and scaling machines take people out of unsupported headings. Wearables track location, heat and fatigue, and augmented reality shows maintenance steps in the technician’s view. Public gold-mine data on these tools is thin, so judge them by the safety record a supplier can show on a comparable site.

Cons: new tools add training time and new failure points. Small-scale operators work very differently and rarely run fleets or control rooms.

10. Blockchain for gold provenance

Best for: refiners and buyers who must prove where metal came from. LBMA and the WGC ran a Gold Bar Integrity pilot with more than 30 participants across 13 locations and 16 completed test cases, where provenance data followed each refined product from its source (LBMA, September 2022). The pilot also showed that industry standards for terms, disputes and re-entry of material still needed agreement.

Cons: a ledger only records what is put into it. It does not prove the first entry was honest, so audits at the mine stay necessary. For sourcing practice on site, see our guide to sustainable gold mining practice.

Do these technologies change the gold price?

Not directly. The WGC reports that mined output was 3,645 tonnes in 2024, only 4 tonnes more than 2023 and just under the 2018 record of 3,658 tonnes, and that annual output has barely moved since 2018 (WGC, January 2026). It adds that mined supply lags the price by at least six years because projects take so long to build. The USGS shows the same pattern in the United States: the average price rose from US$1,801 per troy ounce in 2021 to an estimated US$3,300 in 2025, while US mine output slid from 187 to an estimated 160 tonnes (USGS, Mineral Commodity Summaries 2026).

Chart

Largest gold-producing countries

Mine production, tonnes, 2025 estimate

Chart

US gold price rose while US mine output fell

Index, 2021 = 100 (2025 values are USGS estimates)

What technology does change is the cost side and which ore counts as ore. A lower cost per tonne moved lowers the cut-off grade, so lower-grade rock can become worth mining. Even so, the WGC counts rising costs as a feature of the industry, driven by royalties and higher fuel, freight and consumables costs, and reports that the average AISC margin hit a record US$3,076 per ounce in the first quarter of 2026 because the price rose faster than costs. For company-level context, see our summary of gold mining industry trends and the page on Canadian gold producers.

Technology also does little for the discovery problem. Finding the next mine still takes drilling, permits and many years, a story that goes back to the 1896 Canadian gold rush. For wider tool coverage, our overview of advanced mining technology goes beyond gold.

How to pick a technology for your site

1. Write down your largest cost line per tonne: haulage, power, milling or labour. Start where the money is.
2. Check the ore. Sorting and processing choices depend on how the gold sits in the rock, so run test work on your own sample.
3. Ask each supplier for results from a site with similar ore, scale and climate, and for the date of the data.
4. Count the whole cost: network, training, spares and the staff time to act on the data.
5. Start small. A pilot on one bench, one stope or one stockpile shows the real gain before a site-wide buy.

This page is for information only. It is not investment advice. Prices and costs change; confirm current figures from the sources linked.

Frequently asked questions

What technology is used in gold mining?

Operators use autonomous haul trucks, sensor-based ore sorting, drone and satellite mapping, sensor networks, 3D geological models, processing upgrades, renewable power and safety systems. The mix depends on mine size, ore type and country.

Does new technology lower the gold price?

Not directly. Global mine output has been close to flat since 2018 (WGC, 2026), and the price has risen while US output fell (USGS, 2026). Technology mainly changes costs and which ore is worth mining.

How much of gold mining is automated?

There is no standard global figure for how much gold mining is automated. The clearest count is for haulage: Komatsu alone had commissioned 1,000 autonomous ultra-class trucks by April 2026, across all commodities.

How does AI help gold mining?

AI reads patterns in sensor, drill and plant data. In the Cariboo sorting tests, AI advances allowed the finest 6-10 mm fraction to be sorted, with 59% waste rejection and 92% gold recovery in that test.

What is sensor-based ore sorting?

It scans broken rock with sensors such as X-ray transmission and uses air jets to reject waste, so the mill treats less rock. Results depend on how the gold is spread through the rock.

Can satellites or drones find gold?

Neither sees gold itself. Drones map terrain and stockpiles in detail. Satellites map surface clues such as iron stains, clay minerals and faults, which point to places worth sampling.

Is blockchain used to trace gold?

It has been piloted. LBMA and the WGC tested ledger-based provenance tracking with more than 30 companies in 2022, and further governance work was still needed afterwards (LBMA, 2022).

Why do gold mining costs keep rising despite new technology?

The WGC links the rise to royalties and higher fuel, freight and consumables costs. Its Q1 2026 average AISC of US$1,785 per ounce was 16% above a year earlier.

Does technology make gold mining safer?

Barrick’s then chief executive Mark Bristow said in July 2025 that automated haulage reduces operator exposure to hazards. That is a company statement, not independent injury data.

Can small-scale miners use these technologies?

Mostly not the large-fleet tools. Lower-cost options such as sample testing, drone or satellite mapping and basic sensors are more realistic, and permission and a licence are required before any prospecting, and mercury must never be used.