I do a search for site:www.domain.com for some of my clients to check
how many pages are indexed, but it's coming up with insanely different
results.

For example, one site I check on Google's search engine has 504 pages,
API returns 87.

Another site returns 330 on the search enging, API returns 1.

Why the dramatic discrepancy?

I don't think I've ever seen a result exactly the same from the search
engine and the API.

I have the same problem. I use API to retrieva the first 100 links, but
they are definitely not the same with the real ones returned by google
search engine.

I have seen the same issue.  I am just looking at estimated total
counts.  I was able to get closer search results by ensuring that all
the restrictions were set the exact same way in the API, but there are
still discrepancies in reference counts.  I wonder if the API is being
supported on a different cluster of machines to the real google?

One glaring issue is if you restrict search by language, and there are
no results, then the API returns all results, and there is no way to
tell this other than detecting that the language on the returned pages
is different that specified.

I'm seeing the same discrepancy in results as well.  Here's a search I
just ran:
term: "widget"
via web site: about 3,290,000 results
via api: about 526,000 results

Just a small variance there...
I also think that the API searches through a different cluster then the
web clusters.  Unfortunatly I have a feeling that Google will just say
"well, it's beta!" if asked about this.

And I think Google has shown what they think of the developers using
the API by the complete lack of responses.

At this point I'm giving up on GoogleAPI as a viable tool, since it's
obviously never leaving it's infancy, even though it's been around
longer than the other APIs, it's still the least reliable.

At least Yahoo's support staff respond, and are helpful. You'd think
that somebody would look at the messages, but there are some here that
are well over 3 weeks old and not replied to at all.

Let's hope that Google spend a little more time on the Developer tools
and less time on their April Fools jokes in the future.

I've been experiencing the same problem for weeks (but it seems to have
been going on for much longer), it appears that the API has access to
an index of around 1.6 billion pages compared to google's 8 billion
(does anyone have a more accurate figure?).  I emailed their tech
support over a month ago and I got a reply saying that my email had
been forwarded to the technical team, but I've had no further replies.

The API also seems to return an inconsistent number of pages, for
example running the same query twice gave:

The query "reactive oxygen species" returned about 215000 results.
The query "reactive oxygen species" returned about 137000 results.

If anyone has any ideas of how to reduce any of these inconsistencies
please let me know!

After 3 weeks I finally got a response from google, and it's pretty
much what was expected:

===============================
As you may know, the total number of results displayed when you search
on
Google is only an estimate, not an exact count. Any discrepancies you
notice between a Google search using the Google Web APIs and a Google
search on Google.com are likely the result of ongoing changes to our
index
and/or variations in the estimated number of results.

Regards,
The Google Team
===============================

I am finding the same unreliable results with the api. I have created a
new word and exactly 1 listing appears in a google.com search.

Since this word appeared in google.com I have been tracking it's debut
using the google web api. It has been over 100 days and still no sign
of it. I suspect the inconsistences are more to do with this api
connecting to another caching server other than google.com.

So don't hold your breathe it, looks as if we are entitled to periodic
updates only and not realtime data.

What now?
Querying google.com via HTTP and extracting the results number from a
large chunk of HTML?