#232 Google is "humanizing" the search results to protect itself against the AI storm
When I started in SEO, I learned a ton in forums. In the mid-2000s, social networks became the digital town square and a shift to a closed web.
Now, Google launched a system called Hidden Gems that boosts content in forums or blog posts that have been helpful for a while but don’t have a chance to rank.
My interpretation: Hidden Gem content shows good user engagement but lives on a domain that lacks authority or topical relevance. Similar to the Brand Boost for brand queries like “amazon books”, Google amps up the origin signal.
Some of the biggest benefactors are Reddit, Quora and many other web communities. As I wrote in *ELI5: brand combination searches:* Google’s rank boost for Reddit and other forums was not the result of a Core update but a separate effort. Hidden Gems rolled out in May instead of September as part of the HCU, as first announced.
Instead of having to dig through layers of granite, users can find gems prominently in Search. The question is: why? Besides some users appending “reddit” to their search queries, a big reason is Hidden Gems are not created by AI. Google knows that synthetic diamonds in the form of AI content are cheap to make and about to flood the market. Even fake diamonds, content that’s posted by a human by created by AI, is less likely to come up in well-moderated forums.
But the biggest benefit of forums is user verification.
User verification = antidote to AI commodity content
One big problem AI poses to Google - and a potential reason why Google didn’t release its LLMs sooner even though it invented the transformer - is the disruption of unit economics in the Google economy.
Pre-LLMs, creating content posed healthy friction. Companies paid writers who needed time to create content. Even billions in SEO investment (one can dream) couldn’t bypass time constraints. Now, LLMs create content at a fraction of the cost and time. The results aren’t as good as humans (yet), but good enough for many use cases.
However, real value comes from scarcity, as I explained in the last Memo. AI removes critical friction in the content creation process that keeps Google’s value as aggregator. User verification adds that friction back. If Google prioritizes content that’s guaranteed to be created by humans, it can reduce the negative impact of AI and keep its profitable cash cow alive.
Smart: instead of figuring out how to verify users themselves or bring rel=author back, Google leverages the verification process of user-generated platforms like Reddit and started showing follower stats for authors in the SERPs as a way to highlight credibility. The next step should be to show verification badges and rank content from verified users higher in classic Blue Links and the new Perspectives tab.
All big platforms (Meta, LinkedIn, X) started to verify their users to protect themselves against the crushing wave of AI bots and content. When I recently went through the LinkedIn verification process, Clear, the same company that speeds up security checks at airports, asked me for the ultimate form of identification: my passport. Like showing ID in bars and restaurants when we want to drink alcohol, we might soon have to verify being human on social platforms.
Forums vs. social networks vs. LLMs
There is an interesting tension between gaining value for Search and being disrupted by LLMs. On one hand, forums are sources of human content in Search. On the other hand, a lot of answers people traditionally get in forums can now be given by LLMs.
What’s the value of a forum when an LLM can give a faster and maybe better answer? The conversation itself. Reddit and Youtube comments are often hilarious and make the experience entertaining. You also find experts in niche-focused communities. Most of all, the downsides of social media don’t exist in forums.
Fragmentation in the form of niche forums can be a powerful antiviral drug against doom scrolling, anger boosting and separation because forums are finite, topic-focused and slow.1 X’s vision of being the digital town square worked when town squares brought villages together, but connecting billions of people creates polarization, mental health issues and feelings of loneliness. A true town square is local.
A strong example of the disruption of forums through LLMs is dev community StackOverflow. Internal data shows total traffic down by -28% and number of questions down by -56%.2 The company came out with a statement blaming a tracking issue and correcting the data for traffic being down only -5%. However, the timings don’t fit with the tracking changes and don’t explain the gradual decline of votes and posts that started in 2018.3
StackOverflow is also a perfect example of an aggregator turning into an integrator: its distribution advantage from having lots of content is fading, but content scale offers a data advantage to train a proprietary LLM - which they do. Overflow AI leverages generative AI to help searchers find faster answers, integrate with communication tools and allow companies to build internal StackOverflows.
An open question is how to compensate contributors whose answers are the basis for AI assistants. Getty Images, for example, shares revenue from AI content with photographers.
Funny enough, stackoverflow.com gained lots of SEO visibility since Hidden Gems launched.
In 2017, I wrote, “Controlled online communities are the most underestimated tool for growth and product development in the startup world.” I was neither completely right nor wrong. Had you listened to me and invested in communities, you’d have a massive head start against everyone else who’s now investing in communities. But I never saw coming that the reason would be Google pushing forums again.
You can leverage Hidden Gems for Growth in 3 ways:
1/ The fast food approach to capitalizing on Hidden Gems is by collaborating with the gems through paid influencer campaigns, just like on social networks.
2/ The casual food version is creating (or buying) a profile and gaining more visibility in forums yourself.
Great example: once a year, Tim Soulo from ahrefs asks members of the BigSEO subreddit for product feedback.
You can leverage SEO to prioritize community engagement. For years, SEO used Reddit to mine keyword ideas. Now, we can analyze Reddit’s keywords to find subreddits and threads to engage with. Like SEO or drilling for diamonds, building up verified and trusted profiles takes time.
3/ The fine dining approach is starting your own forum or community.
The Hidden Gems system boosted the SEO visibility of many forums:
Building communities is resource-intensive. You need moderators, infrastructure, clear rules and value. But on the other end is the chance to farm gems.
Gemini was supposed to bring balance to the force but instead left it in darkness. The ChatGPT competitor was supposed to be Google’s return as leader in the AI game. Days later, Google is still playing catchup. Gemini is only slightly better than ChatGPT and not the step change we expected. Bard with Gemini still makes factual mistakes.4
Relevant for SEOs: Gemini improves the quality and speed (+40%5) of SGE, which will make it much more suitable for a roll-out.
Pichai: We’re already experimenting with it in the search generative experience, and as we are experimenting with it, it’s driving improvements across the board. We think about Gemini as foundational — it will work across all our products. Search is no different.
One of the things search has been pushing hard on is multimodality in general. But today they have had to do all the hard work of making search multimodal. Gemini as a foundational model gives them that capability [natively], so I think that’s an area where they will innovate.6
Google hid a little gem in the presentation and staged a Gemini demo that went viral.7 Messing up demos and losing credibility is the opposite of what the company needs. It reminds me of the botched Bard announcement in February, which had a screenshot of a hallucinated answer in the promo video.
When SGE rolls out and if it’s sustainable for Google and users, it could divide and conquer with Hidden Gem content. AI Snapshots that answer questions directly and Hidden Gems add the human sprinkle.
Meanwhile, ever-second-to-Google-search-engine Bing quietly launched a promising experiment: Deep Search.8 With the help of ChatGPT4, Bing disambiguates (clarifies) searches and expands queries to dig for better content and surface better answers. It’s much slower than regular Search (answers can take 30 seconds), but in combination with Bing Copilot and classic Search, it might be a powerful addition. Sounds like a gem.