If you are learning SEO now, the timing is odd but useful. The rules shifted faster in the last year than in the prior ten. People answer questions in ChatGPT, Reddit threads, and YouTube before they click a blue link. Google still matters. It is no longer the only organic surface that matters.
The jaw graph in Search Console
Open Google Search Console on a mature site and you may see a shape people call the jaw: impressions climbing, clicks flat or falling.
AI Overviews sit at the top of many result pages. Google answers the query in the SERP. Your page can still "win" visibility without winning the visit. Some studies cite organic click drops in the 30%+ range when an overview fires. Treat that as directional, not gospel. Your niche and query mix will differ.
That does not mean organic search is dead. It means traffic from Google alone is a shrinking slice of discovery. You plan for Google, YouTube, Reddit, Amazon (if you sell there), LinkedIn search, and whatever LLM your buyer uses this quarter.
Search everywhere optimization (in plain language)
Fifteen years ago, organic almost meant Google. YouTube, Instagram, and TikTok either did not exist or did not behave like search engines.
Today:
- YouTube is the second-largest search engine by usage. Titles, chapters, and descriptions are SEO.
- Reddit ranks for problem-and-recommendation queries. Real participation beats dropping links.
- Amazon is search for shoppers.
- ChatGPT / Perplexity / Gemini surface brands that are cited clearly on the open web.
Search everywhere optimization means you list where your audience actually looks, then you publish native content or answers on those surfaces. Not one blog and a prayer.
The three building blocks (still true, different execution)
Eric Siu groups the durable core into content, citations, and E-E-A-T (experience, expertise, authoritativeness, trustworthiness). The labels are Google's language. The idea is older: prove you know the topic and that others vouch for you.
Content
Content is still the currency. The bar moved. Thin AI pages that say nothing new do not earn links, citations, or repeat visits.
What works in 2025:
- Long-tail queries with real intent. Not "Salesforce" but "how to integrate NetSuite with Salesforce." Not "red shoes" but "best red shoe designs this year." Four-plus word phrases are often lower volume and lower competition, with higher conversion intent.
- Proof in the page: stats, tables, links to primary sources, worked examples.
- Your voice. LLMs can draft fast. They cannot fake ten years of implementation scars. Put those in the post.
You can use ChatGPT, Claude, or agent tools like Manus to brainstorm long-tail lists: give a few seed keywords, describe the business, ask for 50–100 variants, then edit hard before publishing.
Citations (links and more)
"Citations" still means links from other sites for classic SEO. It also means reviews on Google Business Profile, brand mentions in podcasts, and your name on pages LLMs already trust.
You are not buying 10,000 footer links in 2025. You are earning fewer, better references: guest posts that teach, tools people link to, local press, partner pages.
For a deeper taxonomy of link types (and junk to skip), see Types of SEO Backlinks That Actually Move Rankings.
E-E-A-T and "double E-E-A-T"
Google added an extra E for Experience because the web is full of generic explainers. Video, first-person stories, screenshots from real projects, and named authors help.
That carries to YouTube and LinkedIn. A face and a track record beat anonymous "admin" bylines. On a portfolio or services site, that is why I keep author bios and real project links visible, not buried.
AI tools that do most of the grunt work
The transcript claims teams can offload 70–80% of repetitive SEO work if they use agents and automation well. I believe the percentage if "work" means audits, outlines, internal-link suggestions, and first drafts. I do not believe it for strategy, positioning, or anything that touches money and reputation without a human sign-off.
Examples from the source and from practice:
| Task | What tools can help with | What you still own |
|---|---|---|
| Long-tail keyword expansion | LLM chat, Manus-style research agents | Picking keywords that match offers |
| Drafting from style guides | LLM trained on your existing posts | Facts, tone, legal, technical accuracy |
| Internal linking | Agents that crawl your site (Genspark, custom scripts, Cursor/Replit builds) | Which pages matter for revenue |
| Content pruning | Flags thin, cannibalizing, or dead URLs | Delete vs merge vs redirect calls |
| Transcript → outline | YouTube or podcast pipelines | What is worth publishing |
Eric's team built ClickFlow for internal linking, pruning, and scaled drafts. I have not run it on my own sites. The pattern is what matters: compose small tools (even vibe-coded) around Search Console exports and your sitemap, instead of doing everything in spreadsheets.
If you already ship software, a weekend script that proposes internal links from a keyword→URL map is often enough to start.
Surviving AI Overviews without quitting Google
AI Overviews are the synthesized answer block at the top of many Google results. They steal clicks from position one. They also cite sources. Pages that are clear, structured, and trusted sometimes appear in the overview box even when the classic CTR chart looks ugly.
Practical habits:
- Answer the question in the first 100–150 words. Then go deeper. LLMs and overviews scrape the direct block.
- Use real headings (H2, H3). Not cute labels. Match how people ask.
- Add FAQ sections with plain Q/A pairs. Good for humans and for extraction.
- Refresh winners quarterly. Stats age. "2024" in the title hurts in 2026.
- Measure more than clicks. Brand search volume, direct traffic, YouTube subs, form fills from non-Google referrers.
Eric uses a cigar butt metaphor (Warren Buffett): organic Google is declining over a 5–10 year horizon, but there is still margin while others panic. I agree with the timeline, not with neglect. Run Google SEO while you build elsewhere.
Platform notes (where beginners should experiment)
You do not need every channel. Pick two where your buyers already hang out.
YouTube: Hooks in the first 10 seconds, chapters, keyworded titles, transcripts on the site. Repurpose to Shorts if the content fits.
Reddit: Lurk, contribute, cite your post only when it genuinely helps. Spam gets banned and burns the brand.
LinkedIn: Searchable posts and articles for B2B. Native PDF carousels for checklists.
TikTok / Reels / Shorts: Discovery search exists. Hashtags are weak SEO; clear spoken keywords in the first line matter more.
Your website: Still the hub you control. Case studies, pricing, contact, structured data. For a new domain playbook, see How I'd Run SEO on a New Website.
A simple workflow I would run in 2025
- Map search everywhere. List 5 query types your buyer runs. Mark which surface they use (Google, YouTube, Reddit, etc.).
- Build a long-tail list (50+ phrases). LLM-assisted, human-filtered.
- Publish on the site with author, date, internal links, and one external cite per major claim where possible.
- Repurpose once. One article → one YouTube outline or LinkedIn post. Not twelve thin clones.
- Audit internal links monthly. Especially after new service or location pages.
- Prune yearly. Merge cannibalizing posts. Redirect losers. Update posts that rank 4–10 with weak CTR.
For commercial sites in South Africa, pair this with honest service pages and local trust signals. Engineering work lives on Qwabi Engineering; this blog stays the longer-form notes.
FAQ
Is traditional SEO dead?
No. It is spread across more surfaces. Google organic is one leg of the stool.
Should I let AI write all my blog posts?
Use AI for speed on outlines, tables, and meta variants. Publish only what you would sign with your name. Google and users both punish obvious filler.
What is the best long-tail keyword length?
There is no magic word count. Intent matters more than length. "How to X for Y industry in South Africa" beats "SEO tips."
Do AI Overviews always reduce traffic?
Often for informational queries. Commercial and branded queries can still send clicks. Track your own property, not industry averages only.
Manus vs ChatGPT for SEO research?
Both can list keywords and summarize SERPs. Verify anything that affects budget in Ahrefs, Search Console, or manual SERP checks.
Image credits
- Hero and inline photos from Pexels (free license). Downloaded for this post: laptop/analytics and workspace scenes.
Related reading
- Types of SEO Backlinks That Actually Move Rankings
- How I'd Run SEO on a New Website
- What SEO and AEO Are and When to Use Each
- Cross-Platform Libraries and When to Use Each (when the product is an app, not only a site)
If you want help sequencing SEO on a new build or a replatform, get a quote with your domain, market, and which platforms you care about. I will tell you if you need a full program or a smaller audit first.