Conventional PCR and two real‐time PCR (RTi‐PCR) methods were developed and compared using the primer pairs CQULA03F/CQULA03R and CQULA04F/CQULA04R, and TaqMan probe CQULAP1 designed from a species‐specific sequence of the rplJ/rplL ribosomal protein gene, for diagnosis of citrus huanglongbing (HLB) disease in southern China. The specificity and sensitivity of the three protocols for detecting ‘
Candidatus
Liberibacter asiaticus’ in total DNA extracts of midribs collected from infected citrus leaves with symptoms in Guangxi municipality, Jiangxi Province and Zhejiang Province, were tested. Sensitivities using extracted total DNA (measured as copy number, CN per
µ
L of recombinant plasmid solution) were 439·0 (1·30 × 10
5
CN
µ
L
−1
), 4·39 (1·30 × 10
3
CN
µ
L
−1
) and 0·44 fg
µ
L
−1
(1·30 × 10
2
CN
µ
L
−1
) for conventional PCR, TaqMan and SYBR Green I (SGI) RTi‐PCR, respectively. SGI RTi‐PCR was the most sensitive, but its specificity needed to be confirmed by running a melt‐curve assay. The TaqMan RTi‐PCR assay was rapid and had the greatest specificity. Concerning the correlation of PCR detection results with the various HLB symptoms, uneven mottling of leaves had the highest positive rate (96·50%), indicating that leaf mottling was the most reliable symptom for field surveys. Dynamic analysis results from the TaqMan assays showed that the titre (CN) g
−1
citrus tissue of ‘
Ca.
L. asiaticus’ was highest between October and December (threshold cycle (
C
t
) average = 29·3, CN = 3·35 × 10
7
) and lowest between March and May (
C
t
average = 32·0, CN = 5·10 × 10
6
) in 2004 and 2005. The optimized molecular‐based assays should prove useful for presymptom diagnosis of HLB disease, monitoring and identification of ‘
Ca.
L. asiaticus’, and field epidemic regulation.