diff --git a/internal/ast/ast.go b/internal/ast/ast.go
index 9d5cf19d4e..4aa1f4bc2e 100644
--- a/internal/ast/ast.go
+++ b/internal/ast/ast.go
@@ -4,16 +4,135 @@
 package ast
 
 import (
+	"sort"
+	"strings"
+
 	"github.com/hashicorp/hcl/v2"
 	"github.com/hashicorp/terraform/internal/tfdiags"
 )
 
-type AST struct{}
+type AST struct {
+	Nodes []Node
+}
+
+type Node interface {
+	Children() []Node
+	SourceRange() hcl.Range
+	String() string
+}
 
 func FromConfig(files map[string]*hcl.File) (*AST, tfdiags.Diagnostics) {
 	return &AST{}, nil
 }
 
 func WriteAST(ast *AST) (map[string][]byte, tfdiags.Diagnostics) {
-	return nil, nil
+	// Iterate through top-level nodes in the AST and write them back to their
+	// respective files.
+	//
+	// Top-level nodes include their entire content in String(), so we only need
+	// their SourceRange to:
+	// - decide which file they belong to
+	// - determine how many whitespace lines/columns to insert between them
+	var diags tfdiags.Diagnostics
+	if ast == nil || len(ast.Nodes) == 0 {
+		return map[string][]byte{}, diags
+	}
+
+	byFile := map[string][]Node{}
+	for _, n := range ast.Nodes {
+		if n == nil {
+			continue
+		}
+		rng := n.SourceRange()
+		if rng.Filename == "" {
+			// Can't place this node into an output file.
+			continue
+		}
+		byFile[rng.Filename] = append(byFile[rng.Filename], n)
+	}
+
+	if len(byFile) == 0 {
+		return map[string][]byte{}, diags
+	}
+
+	out := make(map[string][]byte, len(byFile))
+
+	// Ensure deterministic output ordering by iterating filenames sorted.
+	filenames := make([]string, 0, len(byFile))
+	for fn := range byFile {
+		filenames = append(filenames, fn)
+	}
+	sort.Strings(filenames)
+
+	for _, filename := range filenames {
+		nodes := byFile[filename]
+		if len(nodes) == 0 {
+			out[filename] = nil
+			continue
+		}
+
+		// Sort by start position (line then column). If ties, sort by end position.
+		sort.SliceStable(nodes, func(i, j int) bool {
+			ai, aj := nodes[i].SourceRange(), nodes[j].SourceRange()
+			if ai.Start.Line != aj.Start.Line {
+				return ai.Start.Line < aj.Start.Line
+			}
+			if ai.Start.Column != aj.Start.Column {
+				return ai.Start.Column < aj.Start.Column
+			}
+			if ai.End.Line != aj.End.Line {
+				return ai.End.Line < aj.End.Line
+			}
+			return ai.End.Column < aj.End.Column
+		})
+
+		var b strings.Builder
+
+		// Pad any initial leading empty lines/columns before the first node.
+		prev := nodes[0].SourceRange()
+		if prev.Start.Line > 1 {
+			b.WriteString(strings.Repeat("\n", prev.Start.Line-1))
+		}
+		if prev.Start.Column > 1 {
+			b.WriteString(strings.Repeat(" ", prev.Start.Column-1))
+		}
+		b.WriteString(nodes[0].String())
+
+		for i := 1; i < len(nodes); i++ {
+			cur := nodes[i].SourceRange()
+
+			lineGap := cur.Start.Line - prev.End.Line
+			switch {
+			case lineGap > 0:
+				b.WriteString(strings.Repeat("\n", lineGap))
+				if cur.Start.Column > 1 {
+					b.WriteString(strings.Repeat(" ", cur.Start.Column-1))
+				}
+			case lineGap == 0:
+				colGap := cur.Start.Column - prev.End.Column
+				if colGap > 0 {
+					b.WriteString(strings.Repeat(" ", colGap))
+				} else {
+					// Overlap or adjacency with unknown intended separator.
+					b.WriteString("\n")
+					if cur.Start.Column > 1 {
+						b.WriteString(strings.Repeat(" ", cur.Start.Column-1))
+					}
+				}
+			default:
+				// Current begins before previous ends; fall back to newline separation.
+				b.WriteString("\n")
+				if cur.Start.Column > 1 {
+					b.WriteString(strings.Repeat(" ", cur.Start.Column-1))
+				}
+			}
+
+			b.WriteString(nodes[i].String())
+			prev = cur
+		}
+
+		out[filename] = []byte(b.String())
+	}
+
+	return out, diags
 }