diff --git a/pkg/storage/bloom/v1/bloom_tokenizer.go b/pkg/storage/bloom/v1/bloom_tokenizer.go
index 96ad1420b303..97dd6b3d0baa 100644
--- a/pkg/storage/bloom/v1/bloom_tokenizer.go
+++ b/pkg/storage/bloom/v1/bloom_tokenizer.go
@@ -112,6 +112,7 @@ func (bt *BloomTokenizer) Populate(swb *SeriesWithBloom, chks Iterator[ChunkRefW
 			chunkSuccessfulInserts int
 			chunkCachedInserts     int
 			chunkCollisionInserts  int
+			chunkBytes             int
 			chk                    = chks.At()
 			itr                    = chk.Itr
 		)
@@ -123,7 +124,7 @@ func (bt *BloomTokenizer) Populate(swb *SeriesWithBloom, chks Iterator[ChunkRefW
 			// raw tokenizer, we could iterate once and just return (prefix, token) pairs from the tokenizer.
 			// Double points for them being different-ln references to the same data.
 			line := itr.Entry().Line
-			sourceBytes += len(line)
+			chunkBytes += len(line)
 			chunkTokenizer := NewPrefixedTokenIter(tokenBuf, prefixLn, bt.lineTokenizer.Tokens(line))
 			for chunkTokenizer.Next() {
 				tok := chunkTokenizer.At()
@@ -174,6 +175,10 @@ func (bt *BloomTokenizer) Populate(swb *SeriesWithBloom, chks Iterator[ChunkRefW
 			}
 
 		}
+
+		// add the recorded chunkbytes to the sourcebytes counter in case we return early via error
+		sourceBytes += chunkBytes
+
 		var es multierror.MultiError
 		if err := itr.Close(); err != nil {
 			es.Add(errors.Wrapf(err, "error closing chunk: %#v", chk.Ref))
@@ -194,6 +199,7 @@ func (bt *BloomTokenizer) Populate(swb *SeriesWithBloom, chks Iterator[ChunkRefW
 		bt.metrics.insertsTotal.WithLabelValues(tokenTypeChunkPrefixed, collisionTypeFalse).Add(float64(chunkSuccessfulInserts))
 		bt.metrics.insertsTotal.WithLabelValues(tokenTypeChunkPrefixed, collisionTypeCache).Add(float64(chunkCachedInserts))
 		bt.metrics.insertsTotal.WithLabelValues(tokenTypeChunkPrefixed, collisionTypeTrue).Add(float64(chunkCollisionInserts))
+		bt.metrics.sourceBytesAdded.Add(float64(chunkBytes))
 	}
 
 	if err := chks.Err(); err != nil {
diff --git a/pkg/storage/bloom/v1/metrics.go b/pkg/storage/bloom/v1/metrics.go
index 6de8c41a791a..22e315e00175 100644
--- a/pkg/storage/bloom/v1/metrics.go
+++ b/pkg/storage/bloom/v1/metrics.go
@@ -17,6 +17,7 @@ type Metrics struct {
 	blockSeriesIterated prometheus.Counter
 	tokensTotal         prometheus.Counter
 	insertsTotal        *prometheus.CounterVec
+	sourceBytesAdded    prometheus.Counter
 
 	blockSize        prometheus.Histogram
 	blockFlushReason *prometheus.CounterVec
@@ -99,6 +100,11 @@ func NewMetrics(r prometheus.Registerer) *Metrics {
 			Name:      "bloom_inserts_total",
 			Help:      "Number of inserts into the bloom filter. collision type may be `false` (no collision), `cache` (found in token cache) or true (found in bloom filter). token_type may be either `raw` (the original ngram) or `chunk_prefixed` (the ngram with the chunk prefix)",
 		}, []string{"token_type", "collision"}),
+		sourceBytesAdded: promauto.With(r).NewCounter(prometheus.CounterOpts{
+			Namespace: constants.Loki,
+			Name:      "bloom_source_bytes_added_total",
+			Help:      "Number of bytes from chunks added to the bloom filter",
+		}),
 
 		blockSize: promauto.With(r).NewHistogram(prometheus.HistogramOpts{
 			Namespace: constants.Loki,