The vast majority of NFL tickets are sold before the start of the season so teams are still claiming sellouts, which while technically correct, gives an impression that isn't reflected in actual people in the seats. I'm not sure how the NFL counts attendance but in NCAA football each school decides on what methodology to use. Some don't count hotdog vendors, some count every person inside the stadium, some just estimate the crowd without bothering with an actual count. It wouldn't surprise me if some pro teams massaged their numbers to appear larger than the actual number of fans who show up.
I have no evidence that if wouldn't surprise me other than actually writing out that it wouldn't surprise me. You'll simply have to take my word that I wouldn't be surprised.
I did some quick calculation here, and it looks like attendance is not lower, in fact, it's slightly higher. 2016 average attendance was 66,586, while this year's average attendance so far is 68,925.
(This doesn't account for lower-attended games towards the end of the season for poor-performing teams; end of the season games often don't sell out for such teams, and these stats may be skewed because we haven't reached the end of the season yet. If I were to guess, by year's end attendance will be slightly lower for 2017.)
What is down is ratings. That is well-documented[1].
Those official attendance figures count tickets issued, not how many fans actually go through the turnstiles. (Teams refuse to release those counts to the public).
You can see massive amounts of empty seats in 4 or 5 NFL stadiums right now. Some cities have extremely loyal fan bases regardless of record (e.g. Green Bay, Pittsburgh), but you can see many seats going completely unsold in Levi's Stadium (49ers).
In fact, Pittsburgh was the reference standard that the 49ers (for one) used as a comparison in selling licenses for their building ("see how WELL those licenses are selling! Don't miss out!").