A coordinate can look precise and be wrong, and it can look coarse and be exactly right, because the digits after the decimal point and the accuracy of the fix are different things that share a screen. This confuses everyone once: a mapping app reports a position to seven decimal places, a resolution of about a centimeter, while the blue dot breathes across a space the size of a house. Both are honest. The decimals describe how finely the coordinate is written. The breathing describes how well the receiver knows where it is. This guide is about the second quantity, accuracy itself: what the official sources actually promise, what eats the promise away in real places, how to read any accuracy claim you are handed, and how the correction technologies, from a better phone to a surveyor's RTK rover, buy the error down rung by rung. If your problem is a phone misbehaving today, our guide to improve GPS accuracy is the repair manual; this article is the theory that tells you what repaired even means.
Home / Coordinates and Location Formats
GPS Coordinates Accuracy: What the Numbers Actually Promise

Photo: Dylan Leagh / Pexels
Read Any GPS Claim: The Accuracy Checklist
Check off each step as you finish it. Your progress stays on this page.
0 of 7 complete
0% done
The Headline Figure, From GPS.gov
Start with the number from the source that runs the system. GPS.gov, the official United States government site for GPS, states the real-world accuracy users should expect this way: GPS-enabled smartphones are typically accurate to within about 4.9 meters, 16 feet, under open sky. Every word of the qualifier matters. Under open sky means the receiver can see a wide hemisphere of satellites with no buildings, canopy, or canyon walls interfering, which is the condition of a field, a highway, or a rooftop, not of a downtown street or a forest trail. Typically means this is the everyday expectation, not a guarantee for your worst moment. And the figure describes the phone's position against the true point on Earth, which is the only accuracy that matters for finding a boundary marker, a geocache, or the right trail junction.
GPS.gov publishes a second figure that is often confused with the first. The GPS signal in space itself, the part the satellite constellation is responsible for, carries a user range error commitment of 2.0 meters or better at 95 percent. That is the space segment's promise about its own signals, before your receiver, your atmosphere, and your surroundings have added their errors on top. Quoting the 2.0 meter figure as the accuracy of a phone fix is the most common exaggeration in consumer GPS writing; the honest consumer number is the 4.9 meter one, and under trees or towers you should expect worse than either.
Accuracy Is Not Precision
Two fixes can disagree about your position in opposite ways, and the distinction names them. Accuracy is closeness to the truth. Precision is closeness to each other. Stand in a city plaza where signals bounce off one glass tower before reaching you and your phone may report positions that agree beautifully, fix after fix within a meter, all of them twelve meters from where you stand: precise, inaccurate, and dangerous to trust, because the stability feels like knowledge. Stand in an open field and the fixes may scatter loosely around your true position, no two alike, centered on the truth: accurate on average, imprecise in detail. The first failure is a bias, a systematic error, and it comes from multipath and geometry. The second is noise, and averaging fixes over time genuinely helps it. Knowing which failure you are looking at tells you whether to wait and average, or to move, because no amount of waiting fixes a bias the environment keeps manufacturing.
The Error Budget: Where the Meters Go
A phone fix is a sum of small thefts. The atmosphere steals first: signals slow as they cross the ionosphere and troposphere, and the receiver's model of that slowing is an estimate, leaving a residue of error that varies with the sun, the weather, and the time of day. Multipath steals second, and downtown it steals the most: a signal that reaches you after bouncing off a building has traveled farther than the straight line, and the receiver, timing the journey, places you farther from the satellite along that bounce, which lands you inside a building or across the street. Satellite geometry steals third: fixes are strong when satellites are spread across the sky and weak when they cluster in one patch, because clustered geometry amplifies every other error, an effect surveyors track as dilution of precision. The receiver itself is the fourth term: antenna quality, chipset, and whether the phone listens on one frequency or two decide how much of the sky's information it can use at all. None of these is a defect. They are the physics of a free, planet-wide positioning service received on a device that also makes phone calls, and the 4.9 meter figure is what remains after all four have taken their share.

Decimal Places: Resolution Is Not Accuracy
The decimal-places table is the most misread table in all of mapping, so here it is computed honestly. One degree of latitude spans about 111,319 meters at the equator, and each added decimal place divides the step by ten:
| Decimal places | Step in latitude | Roughly the size of |
|---|---|---|
| 0 | 111,319 m | A region |
| 1 | 11,132 m | A town |
| 2 | 1,113 m | A neighborhood |
| 3 | 111.3 m | A city block |
| 4 | 11.13 m | A large house |
| 5 | 1.11 m | A doorway |
| 6 | 0.111 m | A hand's width |
| 7 | 0.011 m | A fingernail |
Read the table as resolution, the fineness of the grid the coordinate is written on, never as accuracy. A phone reporting seven decimal places is writing its roughly 5-meter-accurate fix onto a centimeter grid, like reporting your weight to the milligram on a bathroom scale that is off by two kilos. Longitude steps shrink as you leave the equator, since the meridians converge: at 40 degrees north, a degree of longitude spans about 85,276 meters, so the same decimal places resolve finer east-west steps there. Our guide to how many decimal places GPS coordinates need matches decimal counts to real uses; the lesson here is the discipline: keep the decimals your storage and sharing need, and never let them testify about accuracy they do not measure.
The Fine Print: CEP, RMS, and the 95 Percent
Accuracy figures also hide a statistical choice, and reading it is a learnable skill. The same receiver, on the same day, can honestly be described by several different numbers, because "accuracy" needs a rule for summarizing a scatter of fixes. A CEP figure, circular error probable, is the radius that contains half the fixes: a 5 meter CEP means half your fixes land within 5 meters of the truth and half land outside it. A 95 percent figure is the radius that contains nineteen of twenty fixes, and for the same data it is roughly twice the CEP. An RMS figure sits between them in spirit. None of these is the maximum error, which for GPS is unbounded on a bad enough day. This is why the GPS.gov statement in this article is phrased the way careful sources phrase things, a typical figure with its conditions attached, and why a product boasting "accurate to 3 meters" with no statistic named has told you almost nothing: 3 meters as a CEP under open sky and 3 meters at 95 percent downtown are different products wearing the same sentence. When two claims compete, convert them to the same statistic before comparing, or admit that they cannot be compared and discount both.
The Correction Ladder
Each rung of the correction ladder buys down a named error. Dual-frequency receivers, phones and dedicated units listening on both the L1 and L5 signals, measure the ionosphere's delay directly instead of modeling it, and resist multipath better, which is why a modern flagship phone holds a fix downtown that an older phone loses. Satellite-based augmentation, WAAS in North America and its siblings elsewhere, broadcasts corrections for satellite and ionosphere errors from geostationary satellites, tightening ordinary fixes for aviation and marine users. Differential GPS adds a ground reference station at a known point, broadcasting the errors it measures so nearby receivers can subtract the shared part. And at the top, RTK, real-time kinematic positioning, compares the carrier wave itself between a base and a rover and resolves position to centimeter level, the surveyor's tool, bought at the price of a base station or a correction subscription, a short range to that base, and a clear sky to hold the solution. The ladder's pattern is the article in miniature: accuracy is not a property a coordinate has, it is a budget you spend equipment, infrastructure, and sky against.
Reading Any Accuracy Claim
You now own the questions that deflate inflated claims. Accurate under what sky: open, or the urban canyon where the product will actually be used? Accurate by what statistic: a 95 percent figure, an average error, or a best case, which can differ by a factor of three on the same data? Accurate for which device, with which antenna, held how? And accurate against what truth, measured by survey or asserted by marketing? A claim that survives those questions, like the GPS.gov figure this article opened with, complete with its conditions and its source, is a claim you can plan a hike, a survey, or a purchase around. A claim that arrives as a bare number of meters, or worse, a bare count of decimal places, is a claim about the claimant.
Related Guides
This guide does not stand alone. The coordinates guides below take up the questions it leads into:
- GPS Coordinates App for Android: What to Look For, and the Settings That Matter More: Android already knows your coordinates; the right app just shows them well.
- GPS Coordinates With Altitude: Why the Third Number Disagrees With the Map: A GPS position has a third number, and it routinely disagrees with the elevation on signs and maps by tens of meters.
- Plus Code vs what3words: Open vs Proprietary, and Which One to Actually Use: Plus codes and what3words both turn a location into a short code, and they could not be more different underneath.
Frequently Asked Questions
How accurate is a phone GPS, really?
GPS.gov, the official source, puts GPS-enabled smartphones at typically within about 4.9 meters, 16 feet, under open sky. Expect worse among tall buildings, under dense tree canopy, and in steep terrain, where multipath and blocked sky degrade the fix. Expect better from a modern dual-frequency phone in the same conditions, though still not survey grade.
Do more decimal places mean my coordinates are more accurate?
No. Decimal places set resolution, how finely the coordinate is written, not accuracy, how close the fix is to the truth. Seven decimal places resolve about 1.1 centimeters of latitude, but a phone fix carrying them is still accurate to meters. The extra digits describe the grid, not the knowledge.
What is the difference between accuracy and precision in GPS?
Accuracy is closeness to the true position; precision is how tightly repeated fixes agree with each other. Multipath downtown can produce a precise but inaccurate fix, stable and wrong, because the environment biases every measurement the same way. Averaging helps noisy, imprecise fixes; a biased fix has to be fixed by moving or by better signals, not by waiting.
What does the 2.0 meter GPS figure mean, then?
That figure, published by GPS.gov, is the signal-in-space user range error commitment for the satellite constellation itself: the space segment's own error contribution, at 95 percent, before atmospheric effects, multipath, and your receiver add theirs. It is one term in the error budget, not the accuracy of a position on your phone.
How do surveyors get centimeter accuracy from GPS?
With RTK, real-time kinematic positioning: a base receiver at a known point and a rover compare the GPS carrier waves themselves, resolving the rover's position to centimeter level in real time. It needs the base or a correction network, short baselines, and open sky, which is why it is a professional workflow rather than a phone setting.
Photo credits: Dylan Leagh (Pexels); Fran Werner (Pexels). Photos are served from this site; follow the credit to the library page for the original.