CMO playbook & advanced tactics: SEO & content marketing

In February 2025 Chegg sued Google. The complaint argued that AI Overviews were answering students' questions directly on the results page, using material Chegg had paid to produce, and that traffic from non-subscribers had fallen 49% year on year in January 2025. Chegg had already cut about a fifth of its workforce in late 2024, and cut roughly the same share again in May 2025. Revenue slid from around $776 million in 2021 to the $600 million range in 2024, and the stock trades more than 95% below its 2021 high.

The merits of the lawsuit are for lawyers. The mechanism is for you. Chegg's SEOSEOSearch Engine Optimization: the practice of improving your pages' natural (unpaid) rankings in search engine results pages to attract more organic traffic.View full definition → was not broken and its rankings largely held. What changed is what sits above the ranking. A company whose acquisition model rested on one channel discovered a single point of failure nobody had ever been asked to price.

This lesson is about that pricing decision: how much budget stays on the organic line, who you employ to run it, and what breaks first when the call is wrong.

What we are actually talking about

Not whether to do SEO. The arbitration is narrower and harder: what share of demand generation stays on a line whose click yield is falling, which parts of the content operation you keep on payroll, and how you report a channel where rankings hold steady while sessions drop.

Google rolled AI Overviews across US search in May 2024 and kept expanding them through 2025, when the company was describing usage in the billions of monthly users. Pew Research Center, tracking real browsing behaviour in 2025, found people clicked a result on about 8% of visits where an AI summary appeared, against roughly 15% of visits without one. Call it half the click yield on any query that triggers a summary.

The trap is the blended average. One organic number reported monthly hides two opposite trends: informational pages bleeding clicks while comparison, pricing and product pages hold or grow. A CMO watching only the total reacts eighteen months late and cuts the wrong half.

Sub-concept 1: Pricing and reallocating the organic line

Take as given the persistent authority the frameworks lesson defines, and the cluster architecture the application lesson follows through. At your level the question is money.

Split organic sessions into two ledgers: queries that now return an AI answer, and queries that do not. No analytics tool hands you this by default; it takes a query-level export plus a sampled check of live results pages. Then model decay. If 60% of your sessions sit behind informational queries and those lose half their clicks over eighteen months, you lose 30% of the line, and you lose it from the segment that converts worst. That softens the pipelinepipelineAll active sales opportunities across the stages of the sales process, together with their combined potential value and probability of closing.View full definition → hit and does nothing for the conversation with the CFO about a team costing the same and delivering a third less traffic.

The reflex is to cut informational production to zero. That is the expensive move, for reasons that surface two or three quarters later. Informational pages carry the internal links holding up your commercial pages, so their removal drags down the rankings you were trying to protect. They are also the pages AI engines quote, which is where a buyer now forms a shortlist before ever visiting a vendor site: kill them and brand search volume softens later, with no obvious cause on any dashboard. And sales teams are usually the heaviest users of that library, whatever the traffic report says.

There is a second-order cost on the paid side too. Every competitor is reallocating toward the same commercial and transactional keywords at the same time, so auction prices on those terms rise. Google monetises the query either way. Your organic cost per lead advantage over paid narrows exactly when you were counting on it to widen.

The practical fix is to stop running one blended content budget. Run a defence budget on commercial and transactional pages (technical health, freshness, comparison and pricing pages, product-led templates and tools) measured on clicks and pipeline. Run a separate citation budget on original data and durable assets, measured on citations, mentions and brand queries. Declare that second measurement standard when you fund it, not in month nine when someone points at its session count.

Sub-concept 2: Staffing when volume stops being a moatmoatA lasting edge over competitors: a resource, capability or position they cannot easily replicate, letting a firm earn above-average returns over time.View full definition →

A five-person in-house content team in a US market (an editor, two writers, an SEO lead, a designer) runs close to a million dollars a year fully loaded once tools and freelance overflow are counted. That is the line item under threat, and the honest question is not how to defend the headcount but which skills inside it still produce something scarce.

Generative tools pushed the marginal cost of a competent draft near zero, so output volume defends nothing. Google's March 2024 core update, which absorbed the helpful content system into core ranking, came with a stated aim of cutting low-quality, unoriginal results by around 40%. In November 2024 Google issued manual actions under its site reputation abuse policy against large publishers renting out subdomains and directories to third-party affiliate operators, which ended several nine-figure content businesses as going concerns inside a week. The binding constraint moved from production capacity to review capacity, and to whether you own material nobody else can generate.

So the staffing arbitration runs like this. Fewer generalist writers, more analysts, researchers and practitioners who can produce a number or an opinion that did not exist before. One named owner for AI visibility measurement, sitting inside the SEO function rather than as a new team. And a hard look at agency contracts: retainers priced per article are a bad deal when articles are no longer the scarce input, so either renegotiate towards outcomes or bring editorial judgment in-house and buy production capacity outside it.

When a board arrives with an instruction to "use AI to triple content output", the answer that holds up is that volume at near-zero cost consumes the review capacity your citable assets need, and exposes the domain to exactly the enforcement described above.

Sub-concept 3: Distribution and data nobody else has

Salesforce runs this well. Its State of Marketing and State of Sales research surveys several thousand practitioners across dozens of countries, and the resulting figures get quoted by journalists, analysts, competitors and now chatbots for years after publication. Whoever publishes the number an industry wants to quote gets named inside the answer, and gets the links.

Two operational rules follow. First, an anchor asset is not finished at publication: it should already exist as a press angle, an email sequence, a webinar, a set of LinkedIn posts and a sales deck before it ships. Second, watch the gating decision. If your only citable data sits behind a form, AI engines and journalists cite whatever ungated summary exists elsewhere, often a competitor's coverage of your own study. Gate the benchmarking tool or the segment-level cuts; leave the headline numbers open.

Sub-concept 4: What you put in front of the board

Stop reporting raw traffic, and stop assuming rankings and traffic move together, because on any query with an AI answer they no longer do. Report organic-sourced pipeline, cost per organic lead against paid, commercial-intent rank, and click-through rate split by query class.

Add two lines. First, AI citation share: how often your brand appears inside AI Overviews and answers from ChatGPT, Perplexity and Google's AI Mode for your priority queries. Semrush, Ahrefs and a wave of newer visibility trackers now report it, and all of them sell the software they publish research about, which is worth remembering when you read their numbers. Second, a decomposition rule: every decline gets broken into rank loss, click-through loss and query volume loss, because each has a different remedy and the last one is not your fault.

Agree this reporting change while the numbers are still good. Introducing a new metric during a decline reads as excuse-making.

Real-world cases

Case 1: Chegg, and what concentration actually costs. The traffic collapse was the visible event; the organisational failure came earlier. Chegg's shares fell roughly 48% in a single day in May 2023 after management linked slowing new customer growth to ChatGPT, which gave the company almost two years of warning. The response was mainly a product bet, CheggMate, built with OpenAI and announced in 2023, rather than a rebuild of acquisition away from a channel that a single Google interface change could halve. By 2025 the company was litigating and cutting staff in the same quarter. The lesson for a CMO is unglamorous: put a concentration limit on any channel supplying more than half your new customers, and rehearse the scenario where its yield halves in four quarters with no ranking loss at all.

Case 2: Salesforce, and the multi-property problem. Scale here is not query volume but surface area: a corporate blog, Trailhead, extensive help documentation, industry research, plus everything inherited through acquisitions such as Tableau and Slack. Each property can rank for overlapping topics, which means the arbitration is internal before it is competitive. Who owns a topic, which domain gets the definitive page, what happens to an acquired site's authority when it is folded in or left standing. Get this wrong and you fund three teams to compete with each other for the same result, then wonder why none of them holds position. In AI answers the cost compounds, because engines tend to cite one canonical source per claim, and a fragmented estate rarely supplies it.

Case 3: Google, as the counter-example that sets your ceiling. The platform absorbing the click also sells the ad above it, so no reallocation strategy escapes the auction. Two of its policies bound your options: the March 2024 core update against unoriginal content at scale, and the site reputation abuse enforcement that followed. Any plan whose logic is "publish more, faster, cheaper" is a plan to be caught by whichever update lands next. Any plan whose logic is "shift everything to paid" is a plan to buy a rising-price inventory from the same company that changed the rules.

🎬 (Reflect before you read on: if organic clicks halved over the next four quarters with your rankings intact, which line of your budget absorbs the pipeline gap, and how many months of notice would your current reporting give you? If you cannot answer both, you have Chegg's exposure without Chegg's warning.)

CMO action items

• Produce a concentration mapmapUsing software to automate repetitive marketing tasks and campaigns, enabling personalisation at scale across channels like email, web, and social.View full definition → this quarter: new customers by channel, plus the share of organic sessions sitting behind queries that already return an AI answer. Set a written threshold above which a single channel triggers a diversification budget.
• Rebuild the monthly revenue review around organic pipeline, cost per lead against paid, click-through by query class, and AI citation share for your priority terms, with every decline decomposed into rank, click-through and query volume.
• Rebalance the team rather than shrinking it by default: convert at least one writing seat into a research or analyst seat that produces original data, and name a single owner for AI visibility measurement.
• Reopen agency and freelance contracts priced per article, and move the spend toward assets that carry something proprietary.

Common mistakes that kill results

Mistake 1: Treating this as an SEO team problem. The decision about how much organic revenue you are willing to depend on sits with you, not with the person managing rankings. Delegating it means the first honest signal reaches you in a board deck.

Mistake 2: Cutting the whole top of the funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → at once. Informational pages lost their traffic value on many queries, not their structural and citation value. Removing them drops internal link support under the commercial pages you were protecting, and takes you out of the answers where shortlists get formed.

Mistake 3: Running SEO and paid search as separate budgets with separate OKRs. Keywords converting in paid tell you where organic investment will pay back; organic rank data tells you where to taper bids over twelve to eighteen months. Split them and you pay full auction price for terms you already own, or abandon terms you were about to win.

Mistake 4: Measuring content on last-click. A buyer who reads three pages over two months and then requests a demo appears as direct or paid trafficpaid trafficVisitors arriving via paid ads or sponsored placements, where you pay a platform to display your message rather than earning visits organically.View full definition →. Multi-touch attributionMulti-touch attributionA method that distributes conversion credit across all marketing touchpoints in the customer journey, rather than crediting only the first or last interaction.View full definition → is imperfect and still beats a model that systematically undercounts the channel you are deciding whether to fund. Content budgets get cut on this arithmetic every year.

Key takeaways

• Rankings and traffic have decoupled on any query that returns an AI answer, and roughly half the click yield goes with it. Report the two ledgers separately or you will react a year and a half late.
• Chegg's exposure was concentration, not craft. Set an explicit ceiling on how much new business one channel supplies, and rehearse a halving with rankings intact.
• Volume defends nothing now that drafting is close to free. Shift headcount toward original research, review capacity and one named owner of AI visibility.
• Fund a defence budget and a citation budget separately, with different success metrics agreed in advance.
• Reallocating to paid runs into an auction every competitor is entering at the same time, so the organic cost advantage narrows precisely when you need it.