SGE and the End of the Ten Blue Links

For four months now, Google's Search Generative Experience has been live in Search Labs, and we have been running structured tests against client query sets the whole time. If you work in search and have not opted in, do it today, because what is in that experiment is a preview of the biggest change to the results page since ads moved to the top.

The short description: for a growing share of queries, Google now composes an AI answer at the top of the page, synthesized from the web, with a handful of citation cards. The organic results still exist. They are just below a screenful of answer.

What our testing actually shows

We track a few hundred commercially meaningful queries across clients in retail, software, and home goods. Patterns so far, with the caveat that the experiment changes weekly:

• Informational queries are the most disrupted. "How to" and "what is" questions get full AI answers, and the classic top-of-funnel blog post is the collateral damage.
• Transactional and local queries are more protected. Product grids, reviews, and maps still dominate, presumably because that is where the money is.
• Citations are a new competitive surface. Being one of the three sources SGE cites is the new position one, and the content that earns it is not the content that ranked first organically. It skews toward pages with clear structure, specific data, and demonstrable authority.
• Nobody outside Google knows the click-through math yet. Anyone selling you certainty about traffic impact in September 2023 is selling.

The strategic reframe

Here is the way we have started framing it for clients: SEO has been quietly split into two different disciplines, and most content programs are funded against the one that is dying.

The first discipline was answering commodity questions at scale. Ten years of content marketing dogma built enormous blog operations to capture informational queries. SGE eats that layer almost entirely, because Google can now synthesize the commodity answer itself, and frankly the reader is better served. If your content could plausibly be generated, it will be, by the results page itself.

The second discipline is being the source machines cite and the brand people search for by name. Original data, genuine expertise, opinionated points of view, product pages that answer real purchase questions. That layer gets more valuable, not less, because generative answers need credible raw material and because branded demand bypasses the whole fight.

For Northwind, this has meant a hard reallocation this quarter: we cut the generic informational content calendar roughly in half and moved the budget into original research, buying-guide depth on pages that convert, and brand campaigns that create the demand SGE cannot intercept. Search capture is downstream of demand creation. It always was; SGE just removed the arbitrage that let us pretend otherwise.

What to do this quarter

Practical moves, in order: opt into Search Labs and screenshot your money queries monthly, because you need your own record of how presentation evolves. Audit your content by "could a machine have written this," and be brutal. Invest in structured data and clear sourcing, since citation selection visibly rewards it. And start reporting brand search volume alongside organic traffic, because it is about to become the more honest health metric.

The ten blue links had a thirty-year run. Whatever ships out of this experiment, the era where a content farm could rent Google's front page is closing. For clients with real expertise and real brands, that is not a threat. It is the market clearing out the people who never had either.