Google disallows automated access in their TOS, so if you accept their terms you would break them.
That said, I know of no lawsuit from Google against a scraper.
Even Microsoft scraped Google, they powered their search engine Bing with it. They got caught in 2011 red handed :)
There are two options to scrape Google results:
1) Use their API
UPDATE 2020: Google has reprecated previous APIs (again) and has new
prices and new limits. Now
(https://developers.google.com/custom-search/v1/overview) you can
query up to 10k results per day at 1,500 USD per month, more than that
is not permitted and the results are not what they display in normal
searches.
You can issue around 40 requests per hour You are limited to what
they give you, it's not really useful if you want to track ranking
positions or what a real user would see. That's something you are not
allowed to gather.
If you want a higher amount of API requests you need to pay.
60 requests per hour cost 2000 USD per year, more queries require a
custom deal.
2) Scrape the normal result pages
- Here comes the tricky part. It is possible to scrape the normal result pages.
Google does not allow it.
- If you scrape at a rate higher than 8 (updated from 15) keyword requests per hour you risk detection, higher than 10/h (updated from 20) will get you blocked from my experience.
- By using multiple IPs you can up the rate, so with 100 IP addresses you can scrape up to 1000 requests per hour. (24k a day) (updated)
- There is an open source search engine scraper written in PHP at http://scraping.compunect.com
It allows to reliable scrape Google, parses the results properly and manages IP addresses, delays, etc.
So if you can use PHP it's a nice kickstart, otherwise the code will still be useful to learn how it is done.
3) Alternatively use a scraping service (updated)
- Recently a customer of mine had a huge search engine scraping requirement but it was not 'ongoing', it's more like one huge refresh per month.
In this case I could not find a self-made solution that's 'economic'.
I used the service at http://scraping.services instead.
They also provide open source code and so far it's running well (several thousand resultpages per hour during the refreshes)
- The downside is that such a service means that your solution is "bound" to one professional supplier, the upside is that it was a lot cheaper than the other options I evaluated (and faster in our case)
- One option to reduce the dependency on one company is to make two approaches at the same time. Using the scraping service as primary source of data and falling back to a proxy based solution like described at 2) when required.
