Google API and normal SERPS

They don't resemble.

[Hanu]

I run my some handcoded Perl scripts to monitor my rankings on various search engines. Beeing a good person I use Google's API for instead of scraping the the normal serps. I noticed some differences.

First of all, the API seems to show changes before they affect the serps. One or two days. This may be caused simply by the fact that the API uses a different datacenter but it may also be inherent to the API.

The next thing I noticed is this: Imagine the serps for widget city starts with two entries that look as follows:

Buy widget in city!
h t t p://widget-city.com

______Sell or buy used widget in city
______h t t p://widget-city.com/used

Note the indention of the second entry. The Google API returns results like this:

1) h t t p://widget-city.com
...
13) h t t p://widget-city.com/used

The API result no. 13 is promoted to no. 2 in the normal serps just because another page of the same domain is listed at no. 1. Do you know what exactly Google's policies for this promotion are?

[suidas]

Sorry, this isn't an answer, it's a question. Are we sure that Google allows you to use their API to automatically gather positioning info? Is this apple-pie approved?

[Philosopher]

Hmmm...I haven't run into this problem using the API.

May be a silly question, but are you by any chance accidentally doing a phrase search with the API and a default search with Google (one enclosed in quotes and one not).

[Hanu]

Suidas, simply put: yes. You can use the Google API for whatever you want unless you don't share the API key with other people (e.g. publish your interface) and keep the number of queries below 1000 per day.

Philosopher, (nice nick, btw) you startled me for a second so I double checked. Fortunately I'm not doing a phrase search accidentally. One of my searches lists my site at 3, 37, ... and 101 while the equivalent phrase search returns only 39 results.

[Philosopher]

Well, you've got me. As I said, I haven't run into this problem.

What do you have the "filter" option set to?

[kaled]

Others have said, and I think they're right, that a second entry is promoted and indented if it would have appeared on the same page.

Presumably the API does not support this (by default?).

Frankly, I think you should be able to switch off this feature - I find it very annoying.

Kaled.

[shaadi]

Yeah true Hanu, API are not worth doing ranking check anymore, its been a couple of months that I do them by hand - two days of work which can be done in 20 mins flat with API :(

[ogletree]

Check your logs that will tell you more than a checker will.

[Receptional]

I agree with Ogletree - get out of the habit of checking rankings on selected phrases and into the habit of checking traffic from all phrases.

That said, I confess to being more than a little interested in rankings sometimes. It isn't healthy though.

Dixon.

[Hanu]

Philosopher, I'm not sure wether the filter is on or off, since I'm using a wrapper (Perl's WWW::Search::Gooogle). But I don't think it matters. It's not that the API gives a different result set. It returns the same set in a different order.

kaled, I think you hit the nail on the head. I have a page that the API returns at #37. That page gets promoted in the normal serps with 100 or 50 entries per page but it's gone with 10, 20 or 30 entries per pages. Why didn't it occur to me earlier? It's such a simple explanation!

shaadi, ogletree and Receptional: This is the way my clients' brains work. They target obscure phrases and even if these phrases do not generate any traffic (proven by weblog), it makes them happy if they are number one. But for less obscure searches my ranking checks help me not to get lost. You should see the output of my script - it shows you all my sites, all the keyword combos and all ranks for various search engines in a very nice tabular layout. Furthermore,
log analysis only shows you for which keywords your site was found, not the keywords for which it wasn't either found or hit.

[Jetgirl]

I've been using the Google API and I haven't investegated too far, but I am pretty sure it's using a back-up datacenter. The results rarely match 1:1 and are a few days old (aka if I search on "widget" on Tuesday jump up in rank to #3, then finally on Friday if I run the API I'll show up as #3).