Do not rebuild. Roughly 80% of a competent SEO program is unchanged foundation. What changes is measurement, content structure, which topics you choose, and how much effort goes off-site. Restructure your top pages in month one, fix measurement in month two, and shift the content mix in month three.
The most expensive advice circulating right now is that SEO is finished and you should start again.
I understand where it comes from. The traffic numbers are genuinely alarming. Small publishers have lost around 60% of search referral traffic against 22% for large ones, and click-through drops from roughly 15% to about 8% when an AI Overview appears. If you run a content-led business, you have felt that.
But the conclusion does not follow. Nearly everything that makes a page citable by an AI system is something a competent SEO program was already doing. Tearing it up and starting again destroys accumulated authority to chase practices with two years of evidence behind them.
What is required is a pivot, not a rebuild. This is what I would actually change.
What does not change
Start here, because the panic version gets this wrong first.
Technical foundations. Crawlability, speed, rendering, canonical structure. An AI system cannot cite what it cannot retrieve. If anything, rendering matters more now, since AI crawlers are less reliable at executing JavaScript than Googlebot.
Genuine topical depth. The finding that topical authority beats domain size for citations is an argument for doing exactly what good SEO already advises: cover one subject thoroughly rather than many shallowly.
Internal linking and site structure. Hub and spoke architecture serves retrieval as well as it serves ranking.
Earning credibility off-site. Arguably more important now than before, given how much citation appears to depend on being known already.
Conversion. If AI-mediated visitors arrive better informed, as several practitioners report, the page they land on matters more, not less.
That is most of a competent program, unchanged. Anyone telling you to start over is either not doing this work or is selling a rebuild.
What actually changes
Four things, in rough order of how quickly they pay.
1. What you measure
The biggest change, and the one that reframes everything else.
Traffic alone is now a misleading success measure, because visibility and traffic have partly decoupled. You can be more visible and get fewer clicks, and that is not necessarily failure.
Add mention-rate tracking across a fixed prompt set, watch branded search volume as a leading indicator, and start asking inquiries how they found you. Keep traffic in the report, stop treating it as the headline.
2. How pages are structured
The cheapest change with the fastest feedback.
Answer capsule in the first 30% of the page. Headings every 120 to 180 words. At least one table or structured list. An explicit FAQ block. Structured content is cited roughly two to three times more often than prose alone, and around 44.2% of citations come from that opening third.
This is retrofitting, not rewriting. Most existing pages need an hour each.
3. Which topics you choose
The subtler shift, and the one with real strategic consequences.
Purely informational content is now the most exposed category. Informational queries trigger an AI Overview around 39.4% of the time against about 22.2% for commercial ones. A blog answering a definitional question is competing directly with an answer displayed above it.
That does not mean stop publishing informational content. It means stop expecting traffic from it and start valuing it for what it now does, which is establish topical depth and feed citation. Meanwhile shift the share of your effort toward commercial and comparison content, where the click still happens because the reader is evaluating rather than learning.
4. Where the effort goes
If citation is partly post-hoc, off-site presence deserves a bigger slice than it usually gets.
For most programs I would move meaningfully toward off-site work, which for a small firm means directory and profile consistency first, then genuine participation where the topic is discussed. It is slower and less satisfying than publishing, and it is the layer most likely to be the actual constraint.
The 90-Day Pivot Sequence
Concrete order of operations. It is sequenced so each phase produces something you can measure before the next begins.
Days 1 to 30: measure and retrofit
Establish the baseline first. Build the fixed prompt set, run it, record the result. You cannot demonstrate improvement without a starting point, and you will want one in three months when someone asks whether this worked.
Audit for retrievability. Disable JavaScript and check your top pages survive. Check robots.txt against the current AI user agents. Fix anything blocking access, since everything else is theoretical underneath it.
Retrofit your ten most important pages. Answer capsules, heading density, one table where it helps, an FAQ block. Ten pages is roughly a week and it is the highest-return work in the whole sequence.
Fix entity consistency. Organization and Person schema, sameAs links, identical naming across every profile. About a day, and it is embarrassing how often it is left undone.
Days 31 to 60: measurement infrastructure and triage
Set up AI referral tracking in analytics, and separate branded from non-branded queries in Search Console so you can watch demand move.
Triage the content library. Three piles. Pages worth restructuring, because they get traffic or target something commercially useful. Pages worth consolidating, because several thin ones compete for one topic and would be stronger merged. Pages worth leaving alone, because they serve a purpose that is not search.
The consolidation pile is usually larger than people expect and is often the single biggest win available, since it fixes cannibalization and depth at the same time.
Run the prompt set again. Second data point. Do not over-read the difference.
Days 61 to 90: content mix and off-site
Shift the publishing mix toward commercial and comparison topics, while continuing enough informational depth to hold topical authority.
Start the off-site work. Directory and profile consistency is the low-hanging half and can be finished inside the quarter. Editorial and community presence is a permanent activity that starts now and never completes.
Publish one genuinely original thing. Original data, a framework, a benchmark, anything proprietary. The strongest citation configuration combines an answer capsule with proprietary insight, at roughly 34.3%. This is the piece most programs never get to and it is what separates a citable site from a competent one.
Run the prompt set a third time, and write the honest quarterly assessment.
The traffic conversation you will have to have
Somewhere in month two, someone will point out that traffic is flat or down and ask what they are paying for.
Have the answer ready, and make it honest. Traffic to informational content is declining across the board for structural reasons that have nothing to do with your program. The relevant questions are whether qualified inquiries are holding, whether branded search is rising, and whether you appear in AI answers more than you did.
If all three are moving the right way while raw sessions fall, the program is working and the metric is obsolete. If none of them are moving, the metric is fine and the program is not.
What I would resist is quietly changing the reporting to hide the traffic line. Show it, explain it, and put the other three next to it.
When not to pivot
Three situations where I would leave it alone.
Your foundations are broken. If pages are not indexed, or the site is unusable on mobile, AEO is a distraction. Fix the base.
You have no search program at all. Do not start with AEO. Start with the fundamentals, which are 80% of AEO anyway, and add the rest once something is working.
Your business does not come from search. Some businesses run almost entirely on referral and relationships. Pivoting a channel that was never producing is answering a question nobody asked.
The conversation with your existing agency
If you already have an SEO provider, the pivot involves a conversation that can easily go badly, and it is worth going in with a clear idea of what a good answer sounds like.
The wrong opening is asking whether they do AEO. Everyone now says yes, because the alternative is losing the account, and the answer tells you nothing. You will get a confident affirmative and a mention of llms.txt.
A better opening is to ask what has changed in how they measure your program over the last year. A provider who has genuinely engaged with this will have something specific to say about traffic becoming a worse proxy for visibility, and will likely have added something to your reporting. One who has not will describe the same traffic and rankings report they have always sent.
Then ask what they think about the claim that AI citation may be partly post-hoc. You are not testing whether they agree, since it is a contested finding. You are testing whether they have an opinion at all, which tells you whether they read anything beyond marketing blogs.
The third question is the practical one: what proportion of their effort on your account is currently going off-site. If the honest answer is close to zero, that is the gap, and it is a gap in most retainers because on-site work is easier to deliver and easier to show.
What I would not do is fire a provider who is doing solid fundamental work simply because they have not rebranded around AEO. The fundamentals are eighty percent of it. A provider doing genuinely good technical and content work while being sceptical of the newer claims is in a considerably better position than one who has pivoted their marketing language and nothing else.
The situation that does warrant a change is a provider still reporting keyword counts and raw traffic as the headline, with no acknowledgment that the ground has moved, and no interest in discussing it when you raise it. That is not scepticism, it is not having noticed.
One practical suggestion: bring them this article, or the readiness scorecard from earlier in this series, and ask them to score your site. How they respond to an outside framework is informative. A good provider will engage with it, disagree with parts, and tell you which sections they think are wrong. A defensive reaction to a checklist is its own answer.
Frequently asked questions
Is SEO dead?
No. Roughly 80% of what makes content citable by AI systems is standard SEO practice. What has changed is that a portion of demand resolves without a click, so traffic has become a worse proxy for visibility. The work is largely the same and the scoreboard has changed.
Should I stop publishing blog content?
No, but change what you expect from it. Informational content now builds topical depth and feeds citation rather than reliably delivering clicks. Shift some capacity toward commercial and comparison pieces where clicks survive.
How long does the pivot take to show results?
Structural changes can register within weeks because the pages are already indexed. Content mix changes take a quarter. Off-site presence takes several quarters. Expect the first honest read at ninety days and the real one at nine months.
Do I need a separate AEO budget?
Usually not at first. Most of the ninety-day sequence is reallocating existing effort rather than adding spend. The genuinely additive costs are original research, if you commit to it, and off-site presence work.
What if my traffic keeps falling anyway?
Some of it will, and that portion may not return, because the queries were informational and are now answered in place. The correct response is to move effort toward queries where clicks still happen and toward channels you own, such as email. Fighting to recover clicks that structurally no longer exist is the most common way this money gets wasted.
The pivot most businesses need is smaller than the discourse suggests and slower than they would like. Keep the foundations, restructure the pages, change the scoreboard, and start the slow off-site work now rather than when it feels urgent.
If you want a second read on your current program before changing anything, email adphconsulting@gmail.com. ADPH is a US-facing consultancy that delivers its work from the Philippines, and I will tell you honestly when the answer is to change less than you were planning to.