Are the opinion sought quantified? I do understand that averages can only be calculated if you have number so for opinions unless otherwise they are quantified I do not figure out how you can have their averages. May be if you subject the opinions in a Likert Scale, you can rate the opinions and probably you can take the averages of the ratings

If I understand the question correctly, you can aggregate the judgements calculating their geometric mean (not the arithmetic mean, since they are multiplicative, not additive).

You can google for "ahp" and "aggregation of individual judgements" for a list of related publications.

I agree with Pedro. Take the geometric mean and then aggregate the responses. I would also suggest to explore ANP.

Before taking the geometric mean I would suggest to check for consistency of individual judgments by means of the consistency ratio, which should be lower than 10%.

I thank you all for the suggestion, I thought that geometric mean would be appropriate so I took the average using that.

should we take mean of priorities resulting from application of AHP to each respondent's ratings after checking for consistency of each questionnaire ratings?

After having checked for individual consistency, you should make a pairwise comparison matrix in which each element is the geometric means - eventually weighted - of individual pairwise judgments in the same position. From this synthesis matrix you can derive the priorities for all decisonal elements.

Is there any research paper using geometric means as suggested by Elisa Battistoni?

You can look for Saaty T L (2008), Decision making with the analytic hierarchy process, int j services science

Thank you Elisa Battistoni very much for suggesting the paper.