Thursday, August 06, 2009<\/p>\n
\n Earlier this summer \n We hear that Google’s now doing a \n Sites that customize content based on user behavior or clickstream are becoming more common. \n Even though it involves showing different content to different visitors, Google considers \n However, some testing services recommend running 100% of a site’s traffic through the winning \n Running 100% of traffic through one combination for a brief period of time in order to verify \n If the character encoding specified in a page’s HTTP header is different from that specified \n How does Google handle triple-byte UTF-8-encoded international characters in a URL (such as \n Have a question of your own? Join our
\n SMX Advanced<\/a>
\n landed once again in our fair city\u2014Seattle\u2014and it was indeed advanced. I got a number of questions
\n at some Q&A panels that I had to go back and do a little research on. Here, as promised, are
\n answers:\n<\/p>\n
\n better job<\/a>
\n of
\n indexing Flash content<\/a>. If I have a
\n Flash file that pulls in content from an external file and the external file is blocked by
\n robots.txt<\/a>, will that content be indexed in the
\n Flash (which is not<\/em> blocked by robots.txt)? Or will Google not be able to index that
\n content?
\n <\/b>
\n We won’t be able to access that content if it’s in a file that’s disallowed by robots.txt; so even
\n though that content would be visible to humans (via the Flash), search engine crawlers wouldn’t
\n be able to access it. For more details, see our blog post on
\n indexing Flash that loads external resources<\/a>.\n<\/p>\n
\n If a user clicks through to my site from a search results page, can I customize the content of
\n that page or redirect the user based on the terms in their search query? Or is that considered
\n cloaking<\/a>? For example, if someone
\n searches for “vintage cameo pendants” but clicks on my site’s general vintage jewelry page, can
\n I redirect that user to my vintage cameo-specific page since I know that’s what they were
\n searching for?
\n <\/b>
\n If you’re redirecting or returning different content to the user than what Googlebot would see on
\n that URL (for example, based on the google.com referrer or query string), we consider that
\n cloaking. If the searcher decided to click on the ‘vintage jewelry’ result, you should show
\n them the page they clicked on even if you think a different page might be better. You can always
\n link between related pages on your website (that is, link to your ‘vintage jewelry’ page from
\n your ‘vintage cameos’ page and vice versa, so that anyone landing on those pages from any source
\n can cross-navigate); but we don’t believe you should make that decision for the searcher.\n<\/p>\n
\n ethical website testing (such as A\/B or multivariate testing) a legitimate practice that does
\n not violate Google’s
\n guidelines<\/a>. One reason for
\n this is because, while search engines may only see the original content of the page and not the
\n variations, there’s also a percentage of human users who see that same content; so the technique
\n doesn’t specifically target search engines.\n<\/p>\n
\n combination for awhile after an experiment has completed, to verify that conversion rates stay
\n high. How does this fit in with Google’s view of cloaking?\n<\/p>\n
\n your experiment’s results is fine. However, as our
\n article on this subject<\/a>
\n states, “if we find a site running a single non-original combination at 100% for a number of
\n months… we may remove that site from our index.” If you want to confirm the results of your
\n experiment but are worried about “how long is too long,” consider running a follow-up experiment
\n in which you send most of your traffic through your winning combination while still sending a
\n small percentage to the original page as a control.
\n This is what Google recommends<\/a>
\n with its own testing tool, Website Optimizer.\n<\/p>\n
\n in the <meta equiv=\"Content-Type\"><\/code> tag, which one will Google pay attention
\n to?
\n <\/b>
\n We take a look at both of these, and also do a bit of processing\/guessing on our own based on the
\n content of the page. Most major browsers prioritize the encoding specified in the HTTP header over
\n that specified in the HTML, if both are valid but different. However, if you’re aware that they’re
\n different, the best answer is to fix one of them!\n<\/p>\n
\n Chinese or Japanese characters)? These types of URLs break in some applications; is Google able
\n to process them correctly? Does Google understand keywords that are encoded this way\u2014that is,
\n can you understand that www.example.com\/%E9%9D%B4<\/code> is just as relevant to shoes as
\n www.example.com\/shoes<\/code>is?
\n <\/b>
\n We can correctly handle percent-escaped UTF-8 characters in the URL path and in query parameters,
\n and we understand keywords that are encoded in this way. For international characters in a domain
\n name, we recommend using
\n punycode<\/a>
\n rather than percent-encoding, because some older browsers (such as IE6) don’t support non-ASCII
\n domain names.\n<\/p>\n
\n discussion forum<\/a>.\n<\/p>\n