The old SEO bargain is broken. You can still rank, but ranking no longer guarantees anyone will see you, click you, or remember you. AI Overviews, answer engines, social discovery, and LLMs have turned search into a crowded evidence market, and most sites are showing up as anonymous source material instead of destinations.
That is a brutal shift for teams who built their whole strategy around blue links. It also explains why the only durable response is becoming more recognisable, more technically legible, and harder to ignore. Brand authority now does half the job that backlinks used to do on their own, and the technical layer has gone from housekeeping to survival.
Rankings are no longer the whole game
Gerry White’s 3 March 2026 roadmap is useful because it says the quiet part out loud. Search is moving from recommending links to generating answers. Google is still the biggest player, but the old contract, publish good content, win a ten blue links result, collect traffic, is already cracked.
That crack is visible everywhere. AI Overviews sit above the results. Zero-click answers swallow informational queries. People are bouncing into ChatGPT, Grok, Perplexity, Meta surfaces, TikTok and Reddit before they ever reach a classic search page. The job is no longer to own one ranking. It is to be cited, surfaced and remembered across the places where the buyer is actually making decisions.
For South African teams, this matters because traffic volatility is not some distant theory. If you are depending on a handful of high-volume keywords to feed a lead funnel, you are already exposed. The business no longer gets rewarded just for publishing pages that match queries. It gets rewarded for being the source that AI systems trust enough to quote.
Brand authority is the new link building
Several of the experts in the roundup land on the same uncomfortable truth. Mentions matter more. Recognition matters more. A brand that appears repeatedly across trusted sources is easier for both people and machines to select.
Silvia Martin makes the point cleanly. As clicks and organic traffic decline under AI answers and zero-click behaviour, visibility and influence become the real target. She also argues that lower barriers to entry will push more marketers into AI experiments, internal tools, content formats and vibe coding. In other words, the teams that refuse to experiment will not look disciplined, they will look slow.
Anthony Barone pushes the brand point from another angle. His view is that search platforms still care about the same evidence signals, brand, website, social presence, content, authority, qualifications, reputation, and the company site where those things can be verified. He is right to be blunt about it. A lot of businesses are chasing AI noise while their competitors quietly build online equity that will be much harder to dislodge later.
Bengü Sarıca Dinçer adds the other half of the argument. Performance metrics and perception are drifting apart. A site can look fine in analytics and still lose mindshare if nobody recognises the brand, searches for it by name, or sees it repeated on other platforms. That gap is where average brands get buried.
If you want a practical South African frame, think of the difference between a random ecommerce store and a brand people have seen in MyBroadband, on Instagram, in a founder interview, in a Reddit thread, and in product reviews from local buyers. AI systems are learning from that surrounding evidence. So are humans.
Technical SEO is getting more valuable, not less
The idea that technical SEO is less important because AI can write content is backwards. If anything, the web is making technical competence more valuable because low-quality content is now cheap enough for everyone to produce.
Will Kennard’s point is the sharpest one here. AI coding tools are flooding the web with sites that launch with mistakes. That creates more work for people who can actually diagnose broken templates, rendering problems, crawl issues, schema problems, indexation mess, and internal linking failures. The loudest content can be generated by anyone. The technical layer still needs someone who understands how the thing works.
Gerry White is even more direct about where the field is heading. Structured data is moving back to the centre. schema.org, which too many teams have treated as optional garnish, is becoming mission-critical again. AI systems do not just read pages, they consume snippets, entities, citations, metadata and relationships. If your site is unclear about who you are, what you sell, and how pages relate to each other, you are making extraction harder for machines and comprehension harder for users.
He also expects technical SEO to matter more because content is being scraped into training data and overviews. That raises the bar on clean internal linking, content licensing, AI-specific protocols and pages that still function properly without JavaScript. If your product or service page depends on client-side rendering and a prayer, you are creating avoidable risk.
The technical brief is simple. Make the site readable to crawlers, understandable to assistants, and usable without drama.
Human expertise is the content moat
The AI content flood has changed what “quality” means. It is no longer enough to produce another 2,000-word page that repeats the same five subheadings every competitor already has.
Iva Jovanovic says the risk is not that AI replaces SEO, but that businesses spend money on unstable shortcuts while ignoring the durable work. She is right to reject the lazy distinction between SEO and GEO. If a generative engine is surfacing your page, that is still SEO. The naming exercise matters less than the quality of the underlying work.
Yagmur Simsek makes the same point in a more practical way. Topical authority, clean technical foundations, structured data and consistent entity signals are non-negotiable. But technical neatness alone is not enough. The stronger brands pair authority with identity, recognisable experts, clear positioning and a coherent narrative. That is the part AI struggles to fake for long.
Emina Demiri-Watson puts a hard edge on this. She argues that the era of narrow specialisation is fading and full-stack thinking is back. Audience understanding, brand mapping across owned, earned and paid channels, and pre-click messaging aligned with the post-click experience now matter more than obsessing over a single ranking report. She is also right that analytics literacy has to grow beyond the usual SEO toolset. If you cannot explain assisted conversions, repeat visits, engagement depth and cross-channel movement, you are not measuring what the business cares about.
This is where human expertise becomes the moat. Cheap AI pages can rephrase existing material at scale. They cannot demonstrate experience, judgement, field knowledge or accountability. Search systems and users are getting better at spotting that gap.
Measurement has to stop pretending traffic is the only score
The old dashboards are too small for the current problem.
Aleyda Solis expects search to become a hybrid layer inside the buying journey, narrower in pure traffic terms but more important in commercial terms. That is the correct framing. Search is not disappearing. It is being absorbed into a wider decision path where trust, citations and recognition matter before the click ever happens.
Jonathan Moore says the messy middle is now the real battleground. People move between AI answers, feeds, social platforms, apps and assistants before they buy. That means a search report that only counts organic sessions is missing the point. You need proxy metrics and modeled attribution because the journey is fragmented and the final click often lies.
This is especially relevant for South African businesses selling anything with a longer consideration cycle, SaaS, legal services, education, high-ticket retail, B2B services. A lead might start on TikTok, continue in ChatGPT, land on your site via Google, then convert after a WhatsApp follow-up. If your reporting cannot see that chain, your team will keep making bad decisions based on incomplete data.
Brands need to track brand search growth, assisted conversions, repeat visits, qualified lead volume, enquiry quality, and how often the business is being mentioned or cited outside its own site. Traffic still matters. It just stopped being the only useful number.
Collaboration is now part of SEO
SEO used to be the department of the people who handled titles, links and metadata. That version is too small now.
Vanda Pókecz points out a strange but useful contradiction. AI tools and low-code systems are making SEOs more independent because they can build internal tools, monitoring systems, dashboards and prototypes without waiting for engineering. At the same time, fragmentation across search, social, ecommerce and AI surfaces means nobody owns the full journey. That makes collaboration more important, not less.
She is describing the real job for 2026. SEO people need enough product literacy to talk to engineering, enough brand instinct to work with marketing, enough analytical discipline to argue for budget, and enough operational understanding to build tools that solve actual workflow problems. The best practitioners are becoming connective tissue.
Sophie Brannon and Bengü Sarıca Dinçer both land in the same place from different sides. SEO is moving into boardrooms, where stakeholders want business value, not jargon. That only works if the SEO team can explain the trade-offs in plain language and connect them to revenue, retention, pipeline or margin.
For South African teams, the practical version looks like this:
- Product helps clarify what the site actually does.
- Engineering fixes the crawl and rendering mess.
- Content proves expertise instead of padding word count.
- PR and social spread the brand beyond the domain.
- SEO keeps the system coherent.
If the work lives in one channel, it will die in one channel.
The experiments worth running are the boring ones
A lot of AI enthusiasm is smoke. The useful experiments are much more restrained.
Silvia Martin expects more experimentation because the barrier to entry has fallen. That is true, but experimentation only matters if it produces something a business can use. Build internal tools. Test content extraction. Create monitoring scripts. Prototype dashboards. Use LLMs to speed up research and classification, not to replace judgement.
Gerry White predicts that topic authority mapping and entity-based optimisation will replace keyword obsession. That is not a flashy future, but it is the right one. Keyword research still has a place, just not as the whole strategy. The real job is understanding how your site fits into a topic, how your entities relate, and why an AI system should treat you as a source worth citing.
Tom Bourlet’s comments on advertising are a useful reminder that the whole digital stack is shifting at once. Performance Max keeps improving, manual search CPCs keep climbing, Meta changed sharply after its Andromeda rollout in October 2025, and advertisers are being forced into new creative approaches. Search does not exist in isolation from paid media, and SEO no longer gets a free pass to ignore the rest of the funnel.
The winners will not be the teams with the loudest AI demos. They will be the teams that build actual systems, test them properly, and measure whether they changed behaviour.
The practical priority list for 2026
If you strip away the hype, the work comes down to a few things.
First, make the brand impossible to mistake for a generic site. Clear identity, clear authorship, clear proof of competence. Second, fix the technical foundation so crawlers, AI systems and users can extract meaning without friction. Third, stop measuring success as if traffic alone still tells the whole story. Fourth, widen the team around SEO so product, content, development, brand and analytics all pull in the same direction.
That is not a glamorous brief. It is also the only one that survives contact with the current search landscape.
The sites that win from here will not be the ones that publish the most AI-generated pages. They will be the ones that look real, act useful, and are technically clean enough for machines to trust.
