Drawing a Golden Spiral Properly: The Real Curve vs. the Fibonacci-Square Approximation
Search for "golden spiral" and the picture that comes back almost every time is a set of squares, sized 1, 1, 2, 3, 5, 8, 13…, tiled together with a curve of quarter-circle arcs drawn through them. It's a genuinely elegant illustration — and, strictly speaking, it isn't a golden spiral at all. It's a good approximation of one, built from a different and only distantly related piece of mathematics. Getting the distinction right actually makes both pieces more interesting, not less. There are, in fact, three related-but-different mathematical objects hiding behind the single popular phrase, and untangling them is worth doing carefully rather than in passing.
The nested-squares picture, built properly
The classic illustration starts from the Fibonacci-squares identity covered elsewhere on this site: squares with side lengths F(1), F(2), F(3)… F(n) — 1, 1, 2, 3, 5, 8, 13, 21, 34, 55 — tile together, each new square flush against the assembly of all the previous ones, into a rectangle exactly F(n) × F(n+1) in size. Drawing one quarter-circle arc inside each square, sized to match that square exactly and connected end to end across the whole assembly, produces the familiar spiraling curve. It's a real, correct piece of geometry, and it's directly built from the same Fibonacci numbers used throughout this site — nothing about the construction itself is wrong.
Why it's only an approximation, precisely
The trouble is what happens to the curve's shape at each square's boundary. A true smooth curve has a curvature that changes gradually as you move along it. The nested-squares construction doesn't: each arc is a plain quarter-circle with one fixed radius (the side of that particular square), so the curvature is constant within a single square's arc and then jumps abruptly to a new constant value the instant the curve crosses into the next square. A real logarithmic spiral has no such jumps — its curvature changes smoothly and continuously at every point along its length. The nested-squares picture is a sequence of separate circular arcs glued together at their endpoints, not a single smooth curve, and gluing arcs together, however neatly, isn't the same operation as drawing one continuously curving line.
What a true golden spiral actually is
A genuine golden spiral is a logarithmic spiral — a curve of the form r = a·e^(bθ), where r is the distance from the center and θ is the angle swept — with the constant b chosen specifically so the curve's radius grows by exactly a factor of φ for every quarter turn (90°) traveled. Solving for that constant gives b = (2/π)·ln(φ) ≈ 0.306349, and the growth this produces is exact and constant from the very first instant: at θ=0° the radius is 1 (taking a=1 for simplicity), at θ=90° it's exactly φ ≈ 1.618034, at θ=180° it's exactly φ² ≈ 2.618034, at θ=270° it's exactly φ³ ≈ 4.236068, and at a full 360° turn it's exactly φ⁴ ≈ 6.854102. Every single quarter-turn multiplies the radius by precisely φ, with no exceptions and no approximation involved anywhere in the curve.
Plotting the true curve, coordinate by coordinate
The smooth spiral is just as concrete as the nested-squares picture once it's converted to actual coordinates rather than left as an abstract formula. Taking a=1 and stepping through θ in 45° increments: at θ=0°, r=1.0000, giving (x,y)=(1.0000, 0.0000). At θ=90°, r=1.6180, giving (0.0000, 1.6180). At θ=180°, r=2.6180, giving (−2.6180, 0.0000). At θ=270°, r=4.2361, giving (0.0000, −4.2361). At θ=360°, r=6.8541, giving (6.8541, 0.0000) — landing back on the positive x-axis, exactly φ⁴ times farther out than where the curve started. Every one of those radius values is a direct power of φ (φ⁰, φ, φ², φ³, φ⁴), computed from the same PHI constant used throughout this site's calculators, not a separately estimated approximation.
The two curves compared, quarter turn by quarter turn
This is where the "approximation" label earns its precision rather than staying a vague caveat. Compare the radius at each quarter-turn boundary for both curves. The true logarithmic spiral: 1, φ≈1.618, φ²≈2.618, φ³≈4.236, φ⁴≈6.854 — each figure exactly φ times the one before it, from the very first step. The nested-squares construction, read off its actual square sizes: 1, 1, 2, 3, 5, 8, 13… — and the ratio between consecutive values is 1, 2, 1.5, 1.667, 1.6, 1.625, 1.615…, the exact same Fibonacci-ratio sequence that converges toward φ elsewhere on this site, oscillating around the target value rather than sitting on it. The nested-squares spiral's growth factor only approaches φ in the long run; it never equals φ exactly at any single step, and its earliest arcs (a 1-to-1 non-jump, then a doubling) are visibly nowhere close to the true spiral's constant 1.618034 ratio. The two curves do converge toward looking alike as the squares get larger — which is exactly the same convergence phenomenon driving the Fibonacci-to-φ ratio elsewhere on this site — but "converges toward" and "is" are different claims, and the popular illustration only ever earns the first one.
How many squares until the difference stops mattering visually
It's worth being fair to the popular illustration: while it's never exactly a golden spiral, it becomes an extremely close approximation quite quickly, and that speed traces directly back to the Fibonacci-ratio convergence documented at length elsewhere on this site. By the fifth or sixth square, the ratio between consecutive square sizes (8/5=1.6, 13/8=1.625) is already within about 1% of φ; by the ninth or tenth square (55/34≈1.6176, 89/55≈1.6182), the error has fallen below 0.04%, a discrepancy no eye could detect in a printed or on-screen illustration at any ordinary size. So the nested-squares picture isn't a poor approximation — after only a handful of squares it's an excellent one — it's simply not the same mathematical object as the true curve, a distinction that matters for anyone building the spiral programmatically (where "extremely close" and "exact" behave very differently under repeated transformation or magnification) even though it barely matters for a static illustration. even though it barely matters for a static illustration, printed once at a fixed size and never rescaled or recomputed.
The genuinely remarkable property both curves share
None of this is meant to take anything away from why a logarithmic spiral is worth admiring on its own mathematical terms, independent of the golden-ratio-specific version. Every logarithmic spiral, at any growth rate, has the property of self-similarity: scale the whole curve up or down by any amount, and the result is indistinguishable from the original curve, just rotated. It's the same property that reportedly so fascinated the 17th-century mathematician Jacob Bernoulli, discussed elsewhere on this site, that he requested a logarithmic spiral engraved on his tombstone. The golden spiral is simply the specific member of that family tuned to grow by φ per quarter turn rather than some other rate — a distinguishing detail about the growth constant, not about which curves get to count as genuinely self-similar in the first place.
A third, unrelated pattern that also gets called "golden"
To make the naming confusion complete, there's a third pattern in wide circulation that shares the "golden spiral" label despite being built from genuinely different mathematics than either curve above: the discrete, point-by-point pattern this site's own phyllotaxis calculator generates, based on Vogel's model of sunflower seed placement. That pattern doesn't grow by angle-driven exponential expansion at all — its radius grows with the square root of how many points have been placed, and its defining feature is a fixed rotation (the golden angle, ≈137.5°) between successive discrete points, not a continuous curve with a growth-per-turn property. It produces spiral-looking arms as a visual side effect of how the points cluster, covered in depth elsewhere on this site, but it is a materially different mathematical object from both the true logarithmic golden spiral and its nested-squares approximation — three distinct pieces of math, all popularly flattened into the single phrase "golden spiral."
Which one is worth calling "the" golden spiral
If the term is meant precisely, it belongs to the logarithmic spiral with the exact φ-per-quarter-turn growth rate — that's the actual definition, and it's the one with no approximation built into it anywhere. The nested-squares picture earns its popularity honestly: it's easy to draw with a compass and straightedge, it's visually almost indistinguishable from the true curve once a handful of squares have accumulated, and it makes the Fibonacci-to-φ connection immediately visible in a way the exponential formula doesn't. None of that makes it wrong to use as an illustration — it's simply worth knowing, and being able to state precisely, that "almost indistinguishable after a few steps" is a claim about convergence, not about identity.
See the related math directly
The golden ratio calculator generates the golden rectangle each nested square is built from, and the Fibonacci calculator produces the exact square-size sequence behind the popular illustration. The golden spiral & phyllotaxis calculator generates the third, genuinely different pattern discussed above, and the Fibonacci & Lucas convergence reference shows the same ratio-converging-to-φ behavior responsible for the nested-squares spiral only ever approximating, and never exactly matching, the true curve.