Experience will most likely tell you to ignore them. The figures are always too good to be true. But everyone's site is different, so the only way to know for sure is to try it.

I think the experiments are more worthwhile, but in my experience they just side with Google, too. I tried experimenting with displays ads turned off once, and just having text ads, and surprise surprise, every single experiment suggested turning them back on again.
It was the same when I tried doing an experiment without interest based ads. Every single one came back that I should turn them back on

Sometimes it works, sometimes it doesn't. You have to test it.

Of course it will depend on your site, but I have tested many of their recommendations over the years and don't remember a single one having a positive impact, certainly nothing like the levels of improvement they suggest so I don't bother any more.

One thing that I have found worth testing with an experiment occasionally is using different colour text ads ie not always using blended colours but trying a bright colour that stands out instead.

Their suggestions usually haven't worked for me. I generally ignore them now.

Thanks to everyone who has responded

I have implemented some of the recommendations and have seen no increase in earnings whatsoever. If the quoted percentages were true, I would have noticed.

Another reason to distrust what G says as far as I'm concerned.

Another reason to distrust what G says as far as I'm concerned

I fully agree.

Looking over the very long past - how many things have been fashionable only to later fall into disfavour?

I can honestly say that by "doing nothing" over all these years I have never disadvantaged myself with AdSense itself. Google Search is another matter...

Some of the suggestions I've received have been *very* lucrative, and some not, and some were so ridiculous I didn't bother testing them. The point is, you can't go by other people's experiences. They don't have your site, your layout, your audience, your content, your mix of user devices, your location, etc etc etc. The only way you'll ever know is to test it.

Always. Be. Testing.

My account suggests I enable image ads all the time, each time suggesting a solid % increase in earnings. I've re-visited the setting occasionally and every single time my earnings fell rather dramatically. My earnings per click did increase however the huge decrease in actual clicks by far costs me more than the small gain in epc.

Images don't work on my site and I think the message only considers amount earned per click, not any increase or reduction in clicks.

Put it in perspective. G is an ad serving company. What do their employees do? Hype more ads! No-brainer.

However, netmeg is correct. Every once in a while one of g's suggestions just might have a perfect fit for your site ... and you'll never know if you don't play the game (for a period of time).

You can't gain if you don't make change. You can however, as many have noted, lose, so watch closely (minimum six weeks, else it doesn't count) and see what happens.

Remember, ANY change you make will affect both sides of the g equation: search and ads ---- and those two camps do NOT talk to each other.

A few weeks ago I noticed two suggestions like what FarmBoy mentions in this thread.
If I would unblock two "sensitive" categories, it was projected that Category "A" would yield like 20% more income and Category "B" would produce like 30% more.
That was enough of an increase to lower my morals. If I could increase Adsense income along the lines of 40 or 50%, that would be giant to me, so I bit.
The reality was the recommendations were complete and total BS. The numbers they projected were almost certainly just taken from a random number generator.
On the Adsense page where you monitor how different categories perform, after a week the numbers for one category were 0% of total account income/ 0% of account impressions. The other was 0% income/0.1% of impressions.
I get how an ad might be shown and nobody clicks on it, but these numbers suggest that Adsense didn't even have any ad inventory in those categories to show on my site.
The projected vs. experienced difference was so great that I even contacted support about the issue, just to make sure I had implemented things properly.
The reply finally ended up being, oops our bad, thanks for helping to make our program better.
Since disabling those categories, I see new recommendations suggesting I implement them again. On those, the projected increase in earnings has ranged from 5% to 10% to 20%. So, once again, just assume those numbers are totally made up.

In my experience, every site, and its audience is different, and therefore, what works for one may not for another.

I don't doubt that Google's intentions of recommending is meant to help, but what they offer may not be appropriate for your site, although it may be good for another.

As netmeg say, test it. That's what I do, and I ditch anything that doesn't work.

That was enough of an increase to lower my morals. If I could increase Adsense income along the lines of 40 or 50%, that would be giant to me, so I bit.

Well as it turned out, you received the expected result.

However, assuming the AdSense predictions had proven true - how many site visitors would you have likely alienated?

I won't ever alienate my visitors because if my site only had transient traffic, it would be near zero traffic. Something around 90% of my traffic comes from repeat visitors, from links originating on other respected sites, and links embedded in emails, social media, and discussion boards.

90% repeat suggests a forum or community of some sort with little search traffic and those are notoriously finicky to monetize. Adsense offers the option of A/B testing, I strongly suggest any forum type site test, test and re-test.

90% repeat suggests a forum or community of some sort with little search traffic

Oh Duh! Which is precisely what I said. However it's not a forum or community site. The sites are nearly 20 years old. They were only monetised because in retirement I could not possibly afford escalating cost of hosting.

The purpose of the sites was to educate people in a particular genre - not to make money - a claim few others can make in the modern era.

around 90% of my traffic comes from repeat visitors, from links originating on other respected sites, and links embedded in emails, social media, and discussion boards

Over a great many years people have bookmarked the site. Other people [including governments, corporations and educational institutions] linked back to my site as an authority. People corresponding with one another by email embedded links again as an authority - then we had the advent of social media, an extension of email for discussion.

NO it is not, and never has been a forum type site - it is an educational site where after receiving each and every day more questions than I could possibly answer, I established 18 years ago [even before Google was invented] an email 'mutual help' forum. I only participated if someone got technical matters seriously wrong.

In that email forum I never had AdSense or any advertising on it, all it ever carried was a link in the footer to a site page of recommended reading which included Amazon affiliate links which paid all my costs of hosting.

Go for gut instinct and experience. Most advice given by Adsense didn't work on my case, except one time.