Why look back at all?
Our team has spent careers in advertising and media agency groups, through the rise of search, social and programmatic. Each arrived with a rush of tactics that promised to beat the system. Each time, the system changed.
AI answers feel different because the interface is new. But the pattern around them is familiar. Vendors promise placement, brands want a quick win, and the platforms keep changing how they choose sources.
Search, 2003: the shortcut that stopped working
In the early 2000s, ranking well on Google often came down to links and keyword density, and plenty of businesses bought both. In November 2003, Google's Florida update changed the rules overnight, just before the holiday season. Many sites lost their rankings, including legitimate small retailers that had relied on those tactics.
The lesson stuck: tactics that exploit how a system works today are a loan, not an asset. What lasted was content that answered the searcher's question better than anyone else.
Programmatic, 2020: paying for what you couldn't see
Programmatic buying promised efficiency. A 2020 study by ISBA and PwC traced UK programmatic spend from advertisers to publishers. On average, publishers received 51% of what advertisers spent. About 15% couldn't be accounted for at all.
Nobody had hidden that money on purpose. The supply chain was simply too complex for most buyers to see into. The advertisers who did best were the ones who demanded the data.
What does that mean for AI answers?
The three lessons carry over almost exactly:
- Don't buy shortcuts
Promises of "guaranteed #1 in ChatGPT" sound like 2003 link schemes. AI assistants change how they choose sources often, and the same question can get different answers on different days.
- Build what you own
Accurate, current facts on your own pages, and the same facts on the listings and partner pages AI reads. That doesn't disappear when a platform changes its rules.
- Demand the evidence
Ask any provider for the raw AI answers behind their scores: the question, the answer, the date and the sources. If you can't see them, you're in 2020 programmatic again.
What's actually different this time?
Two things, and both raise the stakes. First, there's no organic slot to buy. Some platforms are testing ads, but the answer itself is built from what the assistant can find and trust. Second, buyers often act on the answer without clicking. Pew found people clicked a traditional search result on 8% of visits when an AI summary appeared, against 15% without one. If the summary gets you wrong, you may never see the visit you lost.
Frequently asked questions
Is AI visibility just SEO with a new name?
No. SEO gets you found. AI visibility is about whether the answer names you, and whether it describes you correctly. They overlap, but they're measured differently.
How long until the tactics settle down?
Search took most of a decade. We'd plan on answers changing for years, which is why we measure the same questions over time.
Sources
Social, 2012 to 2014: reach you didn't own
Brands spent years building Facebook followings, assuming they could reach those fans for free. Ogilvy's analysis found a brand page's posts reached about 16% of its fans in 2012. By February 2014 it was around 6%, and about 2% for pages with more than 500,000 likes.
The followers hadn't gone anywhere. The platform had simply decided to show them less. The brands that came through best had treated social as one channel among several, and kept direct relationships with customers.