Early in my career, I worked at Google helping scale the Ads platform. Billions of dollars in advertiser spend moved through systems my teams worked on, and I got a long look at what separated accounts that compounded from accounts that stalled. It was rarely cleverness. The advertisers who won understood how the machine actually made decisions, measured everything, and moved budget the moment the evidence said to. Everyone else optimized folklore: tricks from a conference, habits left over from the last platform shift, dashboards nobody questioned.
I run a growth consultancy now, and over the past year, we built AI search visibility programs for more than a dozen venture and private equity (PE)-backed startups. This transition feels familiar. The front door of search is being rebuilt while most marketing teams keep optimizing the old entrance. Here is the strategy I would build today, on the same principles that worked when the auction was the whole game.
What The Auction Years Taught Me
Three lessons from that era carried over intact.
Platforms reward the signals they can measure, not the effort you put in. Advertisers tried to outspend a weak quality score all the time. It never worked. The accounts that won handed the system unambiguous evidence of relevance and let the machine do the rest.
Budget should follow evidence, and it almost never does. Every account had a line item that existed because it existed the quarter before. The best operators hunted those habits down.
And every platform shift hands value to whoever adapts earliest. Broad match, mobile, automated bidding. Each was declared the end of the channel. Each made a lot of money for the teams that took it seriously a year before their competitors bothered. The pattern repeats because most teams wait for a best-practices deck that only exists after the money has been made.
AI search is the biggest shift of the set, because it does not change how results get ranked. It changes what a result is.
The Game Moved From Ranking To Being The Answer
Your buyers have started asking ChatGPT, Perplexity, Gemini, and Google AI Mode the questions they used to type into a search box. The assistant reads dozens of sources, composes an answer, and names a few brands. The answer is the result now. Your link is optional.
One stat reframed this for me. Semrush studied where ChatGPT’s cited pages rank in traditional search and found that almost 90% of the time, they sit at position 21 or lower for related queries. Read that again. The sources assistants trust are mostly not the pages winning page one. Twenty years of “get to position three” instinct, and the new channel barely glances at that leaderboard.
Client audits keep confirming it. One company we work with had middling Google rankings and a strong footprint on Reddit and in trade publications. Assistants cited them constantly, far out of proportion to their rankings. Another held page-one positions across its category and did not appear in a single AI answer for its core buying queries.
The 6 Signals I’d Build Around
When we audit a company’s AI visibility, six signals explain most of what we find:
- Brand Authority. Assistants favor brands other people mention by name on trusted sources. Your own blog barely counts. Third-party mentions do.
- Content Freshness. Pages untouched for two years get passed over. Visible, dated updates matter more here than they did in classic SEO.
- Entity Recognition. These systems think in entities. They want to know who you are, what you are known for, and which real people at your company are recognized experts on the topic.
- Extraction-Ready Structure. Clear headings, a direct answer near the top of the page, schema where it fits. Danny Sullivan has said structured data supports understanding rather than guaranteeing a win, and that matches what we see. It helps the machine lift what your brand work already earned.
- Cross-Platform Consensus. When Reddit threads, YouTube videos, review sites, and press coverage all describe you the same way, assistants read that agreement as trust. This is the hardest signal for a competitor to copy quickly.
- Earned Media. Coverage you did not pay for and did not write yourself carries outsized weight in what gets cited.
Notice what that list mostly is: Brand and PR work, run with performance marketing discipline. If your AI search plan lives entirely inside your own website, you are working the smallest part of the system.
The 90-Day Sprint I’d Run
We run this with clients as a 90-day sprint in three phases, the same structure I use for 90-day growth audits, pointed at a new target.
Days 1-30: Audit And Foundation
Start with a drill anyone can run this afternoon. Write five prompts a real buyer would ask, questions with money behind them, like “best for [use case]” or “[your product] vs. [competitor].” Run all five in ChatGPT, Perplexity, Gemini, and Google AI Mode. That is 20 runs. Log each one: named, cited with a link, or absent. Show up in fewer than 12 of the 20 and you have a citation problem, which gives the rest of the quarter a clear job. Then make the baseline durable. Tools like Profound, Otterly.AI, and Peec AI track citations across assistants over time, so progress becomes a trendline instead of a feeling. Close the month by picking the two or three channels where the audit shows the biggest gaps.
Days 31-60: Experiment
Run one tightly scoped test per channel, small enough to read in 30 days. Rewrite 10 revenue pages into extraction-friendly formats with direct answers and comparison tables. Build a genuine presence in the two subreddits where your buyers already ask questions. Push for fresh reviews on G2 or whichever platform your category trusts. Pitch two data stories to trade press. Fund all of it by cutting the budget line nobody has questioned in six months. Every marketing org has one.
Days 61-90: Scale And Systematize
Kill what did not move citations. Put real budget behind what did. Build the weekly ritual: the same 20-run drill, the same prompt set, logged against the baseline. Give it a named owner. On most teams that owner is the strategic lead, for reasons I covered when I wrote about building a growth marketing team on a startup budget.
In the Ads years, we lived on impression share. Citation share is the same idea pointed at this channel: of the buying prompts that matter to you, what fraction of answers name you? One sprint we ran this spring moved a B2B client from four of 20 drill runs to 13 of 20 in a single quarter, and the biggest jump came in the two weeks after their first batch of Reddit answers went live.
If this sounds like media buying discipline applied to a channel you cannot buy, that is exactly what it is. You cannot bid your way into an AI answer. You earn the placement with evidence, and evidence takes reps.
Where To Start This Week
You do not need 90 days to start. You need this week.
Run the 20-prompt drill on your own brand. It takes 30 minutes and will change the tone of your next marketing meeting.
Pull channel-by-channel spend and find the line item surviving on habit. Fifteen minutes.
Pick one experiment from the days 31-60 list and commit to running it inside 30 days.
I walked through the full framework, with live examples, in a recent SEJ webinar on building AI visibility in 90 days.
At Google, I watched the advantage go to teams that treated each platform shift as a change in the rules instead of a feature announcement. The rules just changed again. The teams running their first citation audit this week are going to look very smart a year from now.
More Resources:
Featured Image: hmorena/Shutterstock
