SEO in the age of AI Overviews: what actually still works in 2026

By the middle of 2025, AI Overviews and Search Generative Experience had reshaped the top of every search results page in our category. By early 2026, the data is in. Across the client accounts we run, here is what is actually still working in SEO — and what has quietly died.

The myth: “SEO is dead”

It is not. But the SEO that worked in 2018 is dead, and the agencies still selling it are charging you for a corpse.

What is genuinely different in 2026:

• AI Overviews compress 60–80% of informational queries into a generated answer that the user often does not click through.
• “Top of funnel” content — the explainer, the “what is X” — has had its traffic gutted. We have client pages that used to rank #1 for an explainer term and drove 18,000 organic sessions a month; today they drive 1,200.
• Search behavior is bifurcating. Queries that need a definitive provider (booking, buying, comparing) still drive clicks. Queries that need an explanation increasingly do not.

So the question is not whether SEO works. It is which slices of SEO compound now, and which were always borrowed time.

What still works (with data)

Across roughly 40 client accounts we have direct visibility into, four patterns are working consistently in 2026.

1. Bottom-of-funnel terms

The closer a query is to a transaction, the more resilient it is to AI Overviews. A user searching best CRM for solo realtor india is not going to be satisfied by an Overview — they are going to click providers.

We are seeing roughly 70–90% of pre-Overview click-through on bottom-funnel terms across our accounts. The CAC math on these terms has actually improved in many cases, because casual top-funnel traffic dried up but high-intent traffic did not.

Action: Rebalance keyword targets toward commercial and transactional intent. If 70% of your content is “what is” and “how to,” you are building on land that is collapsing.

2. E-E-A-T moats (Experience, Expertise, Authoritativeness, Trust)

Google’s algorithm — even in the Overview era — is increasingly aggressive about elevating content tied to verifiable expertise. We are seeing two specific signals matter more than anything else:

• Author entities. Articles by named, linked humans with on-platform credentials outperform anonymous “Team Author” bylines by 2–4× in stable rankings.
• First-party data. Original research, internal benchmarks, and proprietary data are now disproportionately quoted by AI Overviews (and therefore drive citation traffic). Republishing other people’s research drives almost nothing.

Action: Invest in named author profiles, on-site bios with credentials and links, and at least one piece of original research per quarter. This is hard. That is why it works.

3. Programmatic SEO with real intent

Programmatic — generating many pages from structured data — gets a bad rap because most implementations are spam. But programmatic done with genuine intent matching still works very well, especially in marketplaces, directories, and SaaS.

We have a fintech client whose [city] CA services programmatic stack generates 11,400 organic sessions a month, with a 6% conversion rate to lead. The reason it works: each page has real, useful, locally-specific information — not just keyword-stuffed templates.

Action: If you have structured data (locations, integrations, use-cases, comparisons), build programmatic pages with at least 60% unique, locally-relevant content per page.

4. Branded search velocity

The single most underrated SEO investment in 2026 is non-SEO work: PR, podcasts, LinkedIn founder content, paid social. These do not directly rank you, but they grow branded search volume — and branded search converts at 5–10× non-branded.

In one D2C client account, growing branded search from 800 to 6,200 monthly queries (via founder content) generated more organic revenue than 18 months of pure content SEO.

Action: Treat branded search as a leading indicator and instrument it monthly. If it is not growing, your “SEO problem” is actually a brand problem.

What is dead (and we should stop pretending)

Honest list, based on what we are seeing in account-level data:

• Generic “ultimate guide” content. Done. The Overview eats it. If you are commissioning 4,000-word explainers without an expert hook and original data, you are setting fire to money.
• Keyword density and on-page checklists from 2017. Marginal at best.
• Backlink farming. Has been dying for years. Now it actively hurts because manual reviewers flag inorganic patterns more aggressively.
• Aggressive interlinking. Modest interlinking still helps. The 47-link footer is back to being a negative signal.
• AI-generated longform with no expert review. This is the big one. Pure AI longform ranks for about three months and then collapses. We have watched this cycle play out maybe twenty times in the last eighteen months. Do not believe agencies who claim otherwise.

The new content stack that actually compounds

Here is the shape of content strategy we are deploying for new clients in 2026.

Fewer pieces. Deeper expertise. Original data. Named humans.

Specifically: roughly one expert-led, primary-research-anchored longform per month, plus two to four bottom-funnel comparison or use-case pieces. Programmatic where the dataset justifies it. No “what is X” content unless we can do it materially better than the Overview.

Word count is not the metric. Defensibility is. Ask of every piece you commission: would an AI Overview, given six other articles on this topic, fail to extract the value here? If the answer is no, do not commission it.

A 90-day reset plan

If you are sitting on a content engine designed for the pre-Overview internet, here is the reset we usually run.

Days 1–14: Audit. Pull every page that ranked in the top 20 in 2024. Tag by intent (informational / commercial / transactional). Identify the top-funnel pages losing the most traffic. Identify the bottom-funnel terms you do not own yet.

Days 15–45: Consolidate and redirect. Most content engines have 30–60% redundant pages. Merge, redirect, and strengthen rather than publishing new. This is the cheapest win available right now.

Days 46–90: Build the new pillar. One properly-researched, expert-authored, primary-data piece per pillar topic. Internal links from related content. Author entity wired up. Distribution plan (PR + LinkedIn + newsletter) for launch.

By day 90 you will not have “fixed SEO.” You will have started building the kind of moat that holds for the next five years.

What we get wrong

We are not selling certainty. The truth is that we are six months past a major search inflection and there are still patterns we are guessing about. Some things we have changed our minds on in the last twelve months:

• We were skeptical about programmatic. We were wrong, in specific verticals.
• We assumed schema would matter more than it does. So far, modest impact.
• We thought branded search would matter. It matters more than we thought.

If your current SEO partner has not been visibly updating their thesis in the last six months, that itself is a signal.

---

If you want a 30-minute, no-pitch look at your specific traffic and where the leverage is, write to us. We will tell you honestly whether SEO is the right investment for your stage at all.