The use of the autorefractometer in children (in whom accommodation is more active than older patients) without cycloplegia may underestimate the actual hyperopia and overestimate the actual myopia. Manual retinoscopy is still the most accurate technique to estimate refractive status in children "Network Search", Source:A comparison of manifest refractions, cycloplegic refractions and retinoscopy on the RMA-3000 autorefractometer in children aged 3 to 15 years, Clin Ophthalmol. 2009; 3: 429–431.
Firstly it depends on which auto-refractor you are using and what the purpose of the assessment is. Overall there isn't a big difference and they have been used in large population based studies as they are quick and easy and the mean values agree quite well, and can be useful for screening assessments, but some studies suggest that the accuracy isn't quite good enough for prescribing purposes. They tend to agree that the bigger the refractive error, the greater the variability between measurements.
Below are some links to a few articles on this topic - they cover the issues of accuracy in different devices, and whether they are useful as screening devices.
if eyes are deviated no difference in automated and manual refraction; if eyes are straight there is difference and manual is preferred , for many people it will appear upside down but this is right