Google supports testing and states that running experiments does not harm rankings when done correctly. The risk lies in implementations that look like cloaking.
Rules
- Do not serve Googlebot a different experience based on user agent. Include Googlebot in the normal traffic split.
- Use
rel="canonical"on variant URLs pointing at the original. - Use 302 (temporary) redirects for split tests, never 301.
- Run tests only as long as necessary, then remove the infrastructure.
Alien Road recommends: give variants the canonical tag on day one, not after the test starts. Variant URLs get discovered faster than teams expect, usually through analytics scripts and shared internal links, and once they are indexed you have a duplicate content problem on top of an experiment.