AI Answer Box Keyword Triage
Segment your keywords by intent, quarantine the informational ones, and start teaching the AI
- Difficulty
- Moderate
- Time to result
- ~months to results
- Steps
- 5
- Confidence
- 88%
Levesque's read on the arrival of AI answers at the top of search — which he calls the biggest change since the inception of Google. The framework is a triage: segment your search traffic into transactional, navigational and informational keywords; recognise that informational traffic is the category most likely to be answered directly in the AI box and never reach your site; then plan for the two likely futures (that traffic goes paid, or it simply evaporates) while starting to optimise for a new game — teaching the AI what you do and why you're the best at it.
Origin
Luc Levesque's own analysis, formed at Google I/O 2023 on first seeing the AI answers box in the search results, and pattern-matched against the last decade of Google absorbing verticals — notably its acquisition of ITA and subsequent move into flight and hotel search, which disrupted travel publishers.
Core principles
- 01Every time answers move up the page, entire industries get disrupted.
- 02Google is doing what it has to do — ChatGPT is already taking a large share of daily searches — so this is not reversible.
- 03The historical pattern: free organic traffic tends to migrate to paid, or vanish.
- 04The new optimisation target is not the platform's ranking algorithm but the model's understanding of you.
- 05A platform change is an opportunity — there hasn't been one this big in a long time.
How to run it
- 1
Segment your keywords by intent
Split your search portfolio into three: transactional (e-commerce, purchase intent), navigational (trying to get somewhere specific), and informational (I have a question, I want an answer). Attach current traffic and revenue to each bucket.
- 2
Quarantine the informational bucket as at-risk
Informational keywords are the category most exposed: those are exactly the questions the AI box answers directly in the results. If a meaningful share of your traffic sits there, treat it as a business-continuity issue and model what happens if those clicks disappear.
Watch out This is a broad and often very large bucket — many content businesses are almost entirely informational.
- 3
Model both endgames — paid, or gone
Plan for two outcomes on that at-risk traffic. Either you end up having to pay for what used to be free, or the clicks simply stop arriving because the question was answered in the results. Build the P&L for each and set a trigger for when you'd act.
Pro tip Levesque's timing view: expect this to play out over roughly the next 12 to 24 months if history repeats.
- 4
Switch the optimisation target to the model itself
Stop thinking purely about optimising for the platform's ranking rules and start thinking about teaching the AI what you do and why you're the best in the world at it. Placements are not yet clear, but that is the new game and it's worth being early to.
Pro tip Build your own AI surface too — Lenny's lennybot.com, a GPT trained on his own content, is a template for owning the interface rather than only being cited in someone else's.
- 5
Get reps in with the technology now
Use the tools heavily and personally. The more you use them, the more you understand how good they are and how quickly behaviour will shift. This is the calibration that lets you act ahead of the data.
Watch out Don't dismiss this as 'another change'. Levesque initially framed it as the biggest change in ten years, then revised to the biggest since Google's inception.
In the wild
Google acquired ITA and built flight and hotel search directly into the results page. That was great for users and great for Google — and difficult for the publishers and sites that had been adding value to the ecosystem by ranking in organic results.
→ The travel vertical's organic economics were permanently restructured — Levesque's template for what the AI answer box is about to do to informational search across every vertical.
By mid-2023, Levesque reported that ChatGPT already accounted for 50% of his daily searches rather than Google — which is precisely why he expects Google to react hard with AI answers at the top of the SERP.
→ A concrete leading indicator that the shift is behavioural, not speculative, and that Google's move is defensive and therefore permanent.
Common mistakes
Treating it as an incremental algorithm update
SEO teams have absorbed a decade of tweaks and are conditioned to shrug. This is a platform-level change to where the answer lives, and every prior instance of answers moving up the page disrupted entire industries.
Not knowing your intent mix
If you can't say what share of your organic traffic is informational, you can't size your exposure — and you'll only find out when the traffic is already gone.
Is it for you?
Best for
Growth leaders and content businesses with material organic search traffic who need to size their exposure to AI answers and act before the traffic moves.
Not ideal for
Companies with negligible organic search dependence, or pre-PMF teams with bigger problems.
From the transcript
“there's different types of keywords there's transactional ones like e-commerce keywords and and with purchase intent there's navigational ones where you're trying to get somewhere…”
“I think that last category of keywords is particularly at risk”
“I think the changes will be we'll probably start seeing things shift to paid you have to pay for that free traffic”
“how do you optimize in a world where it's not so much about optimizing for the platform but teaching the AI what you do and…”
“I thought this is actually the biggest changed since since the Inception of Google”
From the episode
Leveraging growth advisors, hiring well, mastering SEO, and honing your craft
Luc Levesque (Shopify, Meta, TripAdvisor)