The short version

  • SEO gets a business found in search engines. An SEO manager owns that work and turns it into revenue. The definition held. The mechanism did not: 68.01% of US Google searches ended without a click in early 2026.
  • SEO still pays off, in a different shape. The same work brings fewer visits with higher intent. Teams that count only visits report a collapse their revenue does not show.
  • Brand gets you into the answer. Depth gets you quoted inside it. Ahrefs studied 75,000 brands: branded web mentions correlate 0.664 with AI Overview visibility, backlink volume only 0.218.

Until recently an SEO status meeting ended with a screenshot. Position one, a rising traffic curve, discussion over. That screenshot proves less today. The page can hold position one while the searcher reads the answer inside Google, decides, and never reaches you.

This is not the end of SEO. It is the end of one metric standing in for a whole job. Below: what the role covers, what it pays, and direct answers to the three questions every marketing lead asks before signing off a budget. Every figure links to a primary source in the list at the end.

What is an SEO manager?

An SEO manager makes a business the answer people find when they search, and turns that visibility into revenue. The work spans the technical side of the site, its structure, the content on it, and the authority the business builds elsewhere. It covers Google’s classic results, AI Overviews, AI Mode, and the assistants people now ask instead of searching.

A specialist executes tasks. A manager owns the outcome and decides what nobody does.

One thing changed in 2026: how the engine presents the answer. For 20 years it returned 10 blue links and the user chose. Now it writes a paragraph and cites three sources. The job held still. The surface moved.

What does an SEO manager do now that most searches end without a click?

Most of the old job survives. Keyword and intent research, technical health, site structure, internal linking, structured data, content briefs, competitor analysis, reporting. All of it became the entry ticket rather than the edge. A slow page never enters the answer at all.

Three things now sit on top, and those three justify the salary.

Write so a machine can quote you

Google documents that AI Overviews and AI Mode both split one question into several related searches, then assemble an answer. Bury your answer under four paragraphs of preamble and the system takes it from whoever put it in the first line.

That skill differs from writing to rank. Usually the SEO manager explains the difference to a content team that spent a decade hearing the opposite.

Measure when there is no click

Search Console shows impressions rising and clicks falling. That is not failure. That is Google answering for you, using your content. Anyone who reports it as a decline reports the wrong thing.

Track three more things now: how often AI answers quote you, how often models name your brand, and where your branded search volume goes.

Work several engines at once

ChatGPT, Perplexity, Claude, Gemini and Google’s AI layers weight different things. Ahrefs measured branded web mentions correlating between 0.66 and 0.71 across ChatGPT, AI Mode and AI Overviews, peaking at 0.709 in AI Mode. You can dominate one and vanish from another. In most companies nobody checks.

Is SEO still as profitable as it used to be?

Not in the same shape. It depends entirely on what you count. Count visits per unit of spend and the ratio really did worsen. Count revenue per unit of spend and the picture looks far better than the headlines suggest.

What did the click loss actually cost?

Start with the losses, because they are real. SparkToro found the share of Google searches producing at least one click fell 9.51 percentage points between 2024 and 2026, a 22.9% relative decline.

Ahrefs found AI Overviews on 20.5% of the 146 million SERPs it analysed. Where one appears, the top-ranking page loses 58% of its clicks. Seer Interactive studied 3,119 search terms across 42 client organisations. Organic CTR on AI Overview queries sat at 0.61% in September 2025, down from 1.76% in June 2024.

Did the fall stop?

Then something happened that nobody predicted. Seer’s April 2026 follow-up ran on a different, much larger sample: 53 brands and 5.47 million queries. CTR on AI Overview queries jumped 85% in two months, from 1.3% in December 2025 to 2.4% in February 2026.

Seer write that the decline “didn’t level, but instead bounced back a little.” They warn against forecasting a recovery off two months. The two samples do not compare, so draw no trend line between them. Say only this: the floor fell hard through 2025, then stopped falling.

Does AI traffic convert better?

Here is the part most of the discussion drops. Traffic that still arrives arrives deeper in the funnel, because the reading and comparing already happened inside the answer.

Adobe Digital Insights watched the sign flip within one year. AI-referred traffic to US retail sites converted 38% worse than everything else in March 2025, 42% better in March 2026, and 54% better in May 2026. Adobe measures over a trillion visits, the largest sample here.

Now the unflattering half. Amsive analysed 54 sites in September 2025: organic at 4.60%, LLM referrals at 4.87%, a gap that failed significance testing at p = 0.794. Meanwhile Search Engine Land published a 13-month read of one agency’s client GA4 data in February 2026 putting LLM conversion near 18%.

Those two sit four times apart. Anyone quoting only the flattering one is selling you something. Amsive’s data also closes mid-2025, exactly when Adobe’s series flips, so their null result probably expired. Safe conclusion: AI-referred traffic now earns at least as much per visit as organic, usually more, and the gap swings hard by vertical.

How much traffic does Google still send?

Ahrefs tracks 107,300 websites. In July 2026 Google accounted for roughly 90% of all search referral traffic. Every AI assistant combined reached about 0.35% of referral traffic, against Google’s 27%.

Move your whole budget off Google because of AI and you are trading on a trend line, not on volume. You will underfund the channel that still delivers nine tenths of the traffic for two years.

So where does the money actually sit?

SEO pays off where intent turns commercial, where your brand is the destination rather than the reference, and where the team counts revenue. It no longer pays as a volume play on informational content that AI now answers alone. That model died. Confusing it with SEO as a whole is the most expensive mistake in marketing budgets today.

Is the new search only about brand, or is there still depth underneath?

Brand gets you considered. Depth gets you quoted. Treat either one as the whole answer and your strategy fails in a predictable way.

What does brand actually correlate with?

Ahrefs analysed 75,000 brands. The three strongest correlations with AI Overview brand visibility all come from outside the site: branded web mentions at 0.664, brand anchors at 0.527, brand search volume at 0.392.

The best link metric in the same study reached only 0.326, Domain Rating, followed by referring domains at 0.295 and raw backlink count at 0.218. Brands in the top web-mention quartile averaged 169 AI Overview mentions, against 14 for the quartile below. In Ahrefs’ follow-up across ChatGPT, AI Mode and AI Overviews, YouTube mentions topped everything at roughly 0.737.

None of that happens by accident. Models learn which brands belong to which topics from how often and in what context they appear across the web. A link only proxies for that, and weakly.

Which page-level signals decide who the model quotes?

From here it looks like SEO became PR. It did not. Brand signals decide whether a model treats you as credible in a category. They never decide which sentence lands in the answer. The page decides that, through one precise, dated, extractable claim that answers what someone actually asked.

Freshness shows this best, and shows why you should never generalise across engines. Ahrefs analysed roughly 17 million cited URLs. Content that AI assistants cite runs about 25% fresher than organic results. Google’s AI Overviews break the pattern in that same dataset and cite content slightly older than organic.

So updating a page moves ChatGPT and Perplexity and does close to nothing for AI Overviews. That distinction separates a strategy from a slogan.

How do the two layers work together?

The brand layer means podcasts, industry press, community presence, YouTube, original research others cite, and one consistent name everywhere. The depth layer means pages that answer narrow questions with specific numbers, full names, dated claims and a real author.

A strong brand with thin pages earns a mention but never a quote. Deep pages under an unknown brand earn the occasional quote and no memory. Most teams fund exactly one of the two.

Will search stay the way it is now?

No. Planning as if it will is the mistake. The direction holds steady enough to build on. The interface does not.

How fast is AI Mode actually growing?

Google announced at I/O in May 2026 that AI Mode passed one billion monthly users, with queries more than doubling every quarter since launch. SparkToro measured only 0.34% of searches moving into AI Mode between January and April 2026.

Enormous in absolute numbers, still tiny as a share of behaviour. Shares move fast from exactly that position.

What could the antitrust ruling change?

Judge Mehta’s remedies took effect on 3 February 2026, without the Chrome divestiture the government wanted. He required annual rebidding of default search contracts instead. The DOJ and its co-plaintiff states cross-appealed the same day. Google filed its opening brief on 22 May 2026. The government filed its combined response and cross-appeal brief on 28 July 2026, captioned “oral argument not yet scheduled.” Google replies by 29 September 2026.

Separately, the technical committee has not finalised the licence terms for how Google shares search data with rivals. Google argues publicly that AI firms should not qualify for index access. Default placement, data sharing and the economics under the whole channel remain open.

What survives every redesign?

Agentic behaviour breaks measurement, not traffic. A background agent completing a task leaves no session, no referrer, no UTM. Conversions will arrive with no origin you can attribute. This already happens at the edges. The teams that cope will build brand-level and incrementality measurement now, before attribution stops making sense.

The economics underneath never move. Somebody wants something, a machine decides who to recommend, and being the recommendation earns money. Every interface since 1998 restated that sentence. Build for the sentence, not the interface.

What does an SEO manager earn?

Published US figures for the same title run from roughly $82,000 to $144,000. Almost all of that spread comes from the sources measuring different things. Read the third column.

What do the salary sources actually say?

Source Reported US figure What it measures
Glassdoor $144,441 Median total pay, base plus additional. Base range shown separately as $84K to $149K
Indeed $89,098 Average base, from 319 salaries in job postings over 36 months
SalaryExpert (ERI) $89,510 Modelled survey average, bonus quoted separately
ZipRecruiter $86,206 Modelled from postings plus third-party data
PayScale $81,910 Average base, 389 self-reported profiles

Figures checked 11 August 2026. Glassdoor reports the only total-pay number here, and the only median, which puts it 60% above the rest. Drop it and four independent sources land within $8,000 of each other. That tells you far more than the headline range.

ZipRecruiter puts the distribution between $70,000 at the 25th percentile and $97,500 at the 75th, with $117,000 at the 90th. In the UK, Robert Half’s 2026 guide gives a £40,500 national midpoint for an SEO and PPC specialist, rising to £55,000 in London on a £45,000 to £77,250 range. That covers a specialist grade, so read it as a floor.

Which skill moves the number?

One skill moves it more than any other. Search For Hire, a recruitment agency in this field, analysed 328,650 job postings from SalaryGuide.com. In June 2026, 20.3% of SEO ads mentioned AI search, GEO or AEO skills, up from 12.2% in December 2025.

Across 13,779 US postings with advertised salaries, those roles showed a median of $117,500 against $97,500, a 20.5% premium. Of the 720 postings naming AI search in the title, 28% were leadership roles, against 15% for standard SEO titles. Read that with the source in mind. A recruiter benefits from a talent-shortage story, and advertised salary is not what companies pay.

Semrush supports the direction. Their 3,900 US Indeed listings, collected November 2025 and published March 2026, put senior leadership titles at 59% of SEO listings, against 15% for SEO Specialist and 10% for SEO Manager. Their senior bucket includes Lead and Executive, so treat 59% as a direction.

Both agree on the shape. AI is absorbing keyword research, brief writing, rank tracking and routine audits. The floor of this profession is disappearing while the ceiling rises. Good news if you decide what matters. Bad news if you only execute.

Should you hire an SEO manager, use an agency, or neither?

Ask one question: does search carry acquisition, or support it?

If organic and AI-referred visibility should produce a meaningful share of pipeline, keep ownership inside the business. The work runs continuously and needs access to product, dev and sales that no external party fully gets.

If search supports the mix, an external team usually wins on economics for the first 12 to 18 months. You buy pattern recognition across many sites instead of a salary, and you learn whether the channel justifies a hire before you commit. That describes most of our search and AI visibility work. When a client’s category does not support the investment, we say so early. That costs less than a retainer proving it slowly.

The third option, neither, holds up more often than agencies admit. Some businesses serve customers who never search before buying. Others compete where AI answers the question completely and sends nobody anywhere. Both should put that budget into demand generation. Any partner unwilling to say so is selling a retainer, not an outcome.

Wherever you land, start the same way, and it costs nothing. Open Search Console, pull 12 months, and find the pages where impressions rose while clicks fell. That gap is your content answering questions you get no credit for. It is also the most reliable list of pages worth rewriting first, because the demand is already proven.

Sources

All figures checked 11 August 2026. Salary data and the Ahrefs traffic tracker update continuously, so treat those two as dated snapshots.

Search behaviour and click loss

Conversion quality of AI-referred traffic

Brand signals, citation and freshness

Platform and regulatory position

Pay and hiring

Dror Feldman is a strategic marketing leader and AI systems builder. He runs Content Managers and works on search and AI visibility for clients in the US, UK and Israel.