For AI Overviews, mostly true: cited pages were statistically indistinguishable from uncited ranking pages on originality (medians 52 vs 55.5, p=.07) across 793 measured AI citations - AIO citation rides on Google rank, not information gain.

In most consumer verticals they dominate: 80% of CPG/retail AI answers cite at least one neutral reviewer or aggregator (Foglift 2026 benchmark - 375 buyer-intent responses, 25 verticals, 5 engines). The exception is tech SaaS, where vendor first-party content is cited 92.7% of the time.

Adding numeric statistics, authoritative quotes and fluent structural edits boosts visibility in generative engines by up to 40%.

Generative engines systematically favor earned media and authoritative third-party coverage over brand-owned and social assets.

Pages with high quote density and numeric stats are cited 2-3x more often than thin content.

When an AI Overview appears, the top-ranking page's CTR drops ~34.5% on informational queries (Ahrefs, 300K keywords); Pew found clicks on traditional results roughly halve when an AI summary is present.

Listicles genuinely earn AI citations: Wix's AI Search Lab found them the #2 cited format for informational intent after articles (~45%), across ChatGPT, AI Mode and Perplexity. Format fit beats format volume.

Perplexity answers synthesize multiple sources rather than a single top-ranked page; Profound's 10-month cross-platform analysis (Aug 2024-Jun 2025) maps how differently each engine picks them.

Structural feature engineering — heading hierarchy, information chunking and visual formatting — independently raises citation rates by 17.3%.

44.2% of LLM citations originate in the first 30% of a document's body text. Lead with the answer.

Branded web mentions are the strongest correlate of AI Overview brand visibility (r~0.39), well above backlinks (r~0.22), across Ahrefs' 75K-brand analysis.

A new causal GEO measurement framework shows why raw mentions are only one input: business impact also depends on query volume, each engine's share of use, whether people notice the mention, and the response under alternative treatment sequences.

LLM citations carry a strong content-recency bias, strongest in AI Overviews; fresh content wins time-sensitive inclusion over older evergreen pages.

Claude and ChatGPT cite from nearly disjoint pools: only ~13% of cited domains overlap (Otterly, 379K citations). Claude skews to company/product domains (64% of citations); Wikipedia is 2.1% and social just 0.9%.

Optimizing high-level features — structural, content and linguistic properties — outperforms token rewriting for citation lift and preserves readability.

Platforms run LLM-as-a-Judge rubrics — helpfulness, reliability, cross-source corroboration — and drop retrieved sources that fail before synthesizing an answer.

In Seer's controlled 123-page test, adding detailed author bylines roughly doubled Bing AI citations period-over-period (vs +34% on control pages); AI Overviews showed no significant lift.

Some assistants (Claude with fetched URLs, Custom GPTs, MCP clients) parse markdown more cleanly than HTML, and .md endpoints reduce token cost when an agent retrieves your page. Direct citation lift is unproven in independent tests.

They help, as multipliers: in a 2M-citation statistical analysis, structured headings, FAQ sections and TLDR/BLUF blocks showed consistent positive effects on citation - on top of content alignment and authority, not as substitutes.

No major generative engine has publicly confirmed reading llms.txt for retrieval or citation as of mid-2026. It is a proposed convention (llmstxt.org), not an announced ranking or ingestion signal.

Schema markup has minimal impact on generative citation across major engines.

Cloaking content specifically for AI crawlers — showing bots material the human page does not contain — risks the same spam classification as classic SEO cloaking. Neither Google nor OpenAI has endorsed the practice.

No major LLM crawler reads llms.txt. It is a signaling gesture, not a technical requirement.

At any moment only ~5% of business buyers are in-market; the other 95% are future buyers who must be reached now to be remembered later.

Long-run profit is maximised near a 46% brand / 54% activation split in B2B, versus 60/40 in B2C.

Brand effects accumulate over 6+ months; measurement windows shorter than two quarters systematically undervalue brand and overvalue activation.

Category-entry-point memory built through broad reach predicts future buying better than narrow retargeting of today's list.

Excess share of voice above share of market predicts subsequent share growth; roughly every 10 points of eSOV maps to about 0.5-1 point of annual share gain.

Ehrenberg-Bass research shows B2B buyers start from a mental shortlist built from category entry points - the cues (motives, situations, occasions) that retrieve brands from memory. Brands grow by linking themselves to more CEPs, not by winning rational arguments.

Typical enterprise purchases involve 6-11 stakeholders; single-contact scoring systematically misreads deal readiness.

Consistent distinctive assets (logo, colour, character, tagline) drive recognition and recall far more reliably than claimed product differentiation.

In Google/CEB/Motista research, B2B purchasers were almost 50% more likely to buy when they saw personal value (career advancement, confidence, pride) in the decision and 8x more likely to pay a premium; only 14% would pay a premium for perceived business-value differences.

Gating trades most of a piece's reach and memory-building for a small list of low-intent form-fills; ungated distribution plus a light capture path usually wins on pipeline.

Contact odds fall off sharply within the first hour of an inbound enquiry; response inside five minutes materially outperforms same-day follow-up.

Buying is looping, not linear: buyers revisit the same jobs repeatedly and contact suppliers throughout, so the single 70% number is misleading.

Average large-site checkouts ask for far more fields than necessary; trimming to the minimum viable set is one of the highest-yield conversion changes available.

Mandatory account creation is a top-cited abandonment driver; guest checkout with a post-purchase account offer captures both revenue and the record.

Shipping cost is the single most cited reason for abandonment; a free-shipping threshold typically outperforms an equivalent-margin discount and lifts average order value.

Frequent deep promotions retrain reference prices and depress baseline sales between promotions, so measured promo ROI overstates true incrementality.

Field studies across large retailers repeatedly tie sub-second load improvements to measurable conversion and revenue gains, with mobile most sensitive.

Triggered commerce email (browse, cart, post-purchase) continues to produce the strongest measured revenue per send of any owned channel.

Consumers encode prices left to right, so $2.99 vs $3.00 shifts demand far more than one cent warrants - the left-digit effect, shown across lab and field evidence (Thomas & Morwitz, JCR 2005).

Conversion gains from review volume rise steeply from zero to a few dozen reviews, then flatten; recency and response quality matter more after that.

Documented average cart abandonment sits around 70% across dozens of studies; most of it is browsing behaviour, and only a minority is fixable friction.

Shoppers rely on image sets to answer objection questions; scale, in-use and detail shots reduce returns as well as lifting conversion.

Nine-ending prices produce measurable demand lifts in field experiments, but the effect is modest and can invert for premium positioning.

Across hundreds of IPA case studies, roughly 60% brand / 40% activation maximises long-term profit growth in consumer categories.

Brand growth comes overwhelmingly from increasing penetration among light and non-buyers, not from deepening loyalty.

Across 751 short-term and 402 long-term elasticities from 56 studies (1960-2008), the average short-term advertising elasticity is 0.12 - but the mean long-term elasticity is 0.24, double the short-term, through carryover.

Creative quality accounts for roughly half of advertising-driven sales variance — a larger share than any single media variable.

Emotionally led campaigns produce roughly twice the long-term business effects of rational ones, and the gap widens with campaign duration.

Across 389 BehaviorScan split-cable TV experiments, increasing budget relative to competitors did not increase sales in general - changing brand, copy, or media strategy did. Standard recall and persuasion pretests did not predict which ads would move sales.

Across nearly 600 IPA Effectiveness Award case studies (1996-2018), creatively awarded campaigns have fallen to their lowest effectiveness in 24 years of analysis and are now no more effective than non-awarded campaigns; short-termism is the main driver.

The IPA's recession evidence review finds brands that maintain share of voice through a downturn recover faster and gain share over cutters - cutting spend mortgages the recovery.

Comparing 4,863 effectiveness award cases (2011-2019), the largest business effects came from 'creatively committed' campaigns: more media spend, more channels, and longer durations behind one creative platform - captured in the Creative Effectiveness Ladder.

Double jeopardy: smaller brands have both fewer buyers and slightly lower loyalty — an empirical law observed across categories and decades.

The ARF's evidence review finds true creative wear-out is less common than assumed: repeated exposure more often builds memory (wear-in) than fatigue, and marketers tire of their ads long before consumers do.

Viewability is a necessary floor, not a measure of attention; attention-time metrics predict brand outcomes far better than viewable-impression counts.

Broad, character-led, 'right-brain' creative correlates with far larger long-term effects than narrow message-dense executions.

Modelling IPA cases with System1 emotional-response data, 'dull' low-emotion TV ads need far more media to match effective creative: roughly 10m pounds extra per UK campaign, scaling to an estimated $228bn of extra spend across US TV to match the market-share growth of the most impactful spots.

In a 100-person within-subject study across four products, moderately personalized AI imagery earned the best ad and product attitudes. At high personalization, rising creepiness outweighed the relevance benefit and neutralized gains across ad attitude, product attitude, and purchase intent.

Share of search leads market share by several months in many categories, giving a cheap, high-frequency brand-health proxy.

Recurring fluent devices and consistent assets build recognition faster and cheaper than continually restarting creative platforms.

In a peer-reviewed proof-of-concept, AI-generated static and video pretest ads scored statistically indistinguishable from the real finished ads they were based on - recall, recognition, brand choice, brand attitude, ad liking.

Identity resets discard accumulated distinctive-asset memory; evolution beats revolution unless the brand carries active negative equity.

In an incentivized experiment with randomized exposure, content, and price, display ads shifted beliefs about price and quality even when the copy revealed neither - and those belief shifts causally drove search and purchase.

Most avoidable churn is set in the first sessions: users who reach a core activation action early retain at multiples of those who do not.

Modest retention improvements compound through the customer base and typically move profit more than equivalent acquisition spend, because retained revenue carries no CAC.

Studying four companies over 16 months, Reinartz & Kumar found only a modest correlation between customer longevity and profitability: long-tenure customers expected discounts, were no cheaper to serve, and were not reliably the best word-of-mouth sources.

LTV curves flatten unpredictably; projections built on the first weeks of a cohort routinely overstate long-run value.

Most loyalty programmes reward behaviour that would have happened anyway; measured incremental effects are small and skew to already-heavy buyers.

The field's anchor framework shows customer experience is cumulative across the entire journey - including social, environmental and competitor touchpoints the firm doesn't control - so single-touchpoint optimization misses the real drivers of satisfaction and loyalty.

The DMA's 2026 Marketer Email Tracker (250 marketers) finds email still central - valued for reach, reliability and measurable impact - but AI-powered inbox prioritization and summarization now mediate what gets seen.

Previously-purchased customers convert on reactivation at materially higher rates and lower cost than cold prospects.

Revenue per additional send declines and list fatigue raises unsubscribe and spam-complaint rates; there is a measurable frequency optimum per segment.

Modelling customer equity directly, Rust, Lemon & Zeithaml show reallocating spend to the highest-leverage driver of customer equity beats across-the-board investment - in their airline application, the concentrated strategy maximized return on marketing.

Independent replications find NPS predicts growth no better than ordinary satisfaction measures, and often worse.

Well-communicated, value-anchored increases typically churn a small single-digit share of subscribers while lifting net revenue.

Stated churn reasons skew heavily toward price because it is socially easy to say; behavioural data usually shows usage decay preceded the cancellation by weeks.

The service-recovery paradox appears only under narrow conditions — first failure, low severity, fast fix — and does not generalise.

Across 25 large field experiments with major U.S. retailers and brokerages, the median confidence interval on ad ROI was over 100 percentage points wide; individual-level sales are so volatile (coefficient of variation ~10) that an informative test often needs 10M+ person-weeks.

Large-scale field experiments show branded search and retargeting capture credit for conversions that would have happened anyway; last-click can overstate paid value by an order of magnitude.

Estimating elasticities and ROI across 288 brands, Shapiro, Hitsch & Tuchman find ad elasticities far smaller than the published literature suggests, negative ROI at the margin for more than 80% of brands, and positive overall ROI for only about a third.

Geo-based holdout experiments give unbiased incrementality estimates without user-level tracking and are the most practical causal method for most advertisers.

Broad reach with modest frequency consistently outperforms narrow high-frequency plans; marginal response per additional exposure declines quickly.

A randomized experiment with 1.6 million Yahoo! users found display ads profitably lifted a major retailer's purchases by 5% - but 93% of the increase happened in brick-and-mortar stores, and 78% came from customers who never clicked the ads.

Across 663 Facebook RCTs described by 5,000+ features, state-of-the-art observational methods (double/debiased ML, propensity matching) miss the experimental lift by a median 62-115% depending on funnel stage - larger than the median lift itself.

A significance test answers whether the experimental means differ, not whether deployment will pay off under future uncertainty. Across 552 advertising experiments, a distribution-aware decision rule substantially reduced regret versus conventional hypothesis testing.

In platform meta-analyses creative explains far more outcome variance than incremental targeting precision, especially in auction environments that already optimise delivery.

Ghost ads - logging the impression a control user would have been served, without serving or paying for it - reproduce RCT-grade measurement at a fraction of the cost of PSA controls, and are precise enough to have become standard practice at Google and for brands like Duracell and Nissan.

MMM and experiments are complements: experiments calibrate priors, MMM allocates across the whole mix. Neither alone is sufficient.

Most retail-media ROAS counts sales that would have happened anyway; even incremental ROAS is so methodology-sensitive that Ovative, Albertsons Media Collective and Kellogg faculty built a framework just to standardize it.

Controlled experiments repeatedly find retargeting incrementality far below reported ROAS, because it reaches users already on a purchase path.

Across 288 brands in many categories, TV ad elasticities come out far smaller than the published literature suggests, and marginal ROI is negative for more than 80% of brands.

Independent audits find a substantial share of programmatic spend disappears into unattributable fees and low-quality inventory before reaching a working impression.

Google open-sourced its MMM (Meridian, January 2025) with Bayesian calibration from experiments built in, and Meta's Robyn is also free - credible MMM now costs analyst time, not a vendor contract.

Using 3.3 million survey responses across 9,596 display campaigns, Goldfarb & Tucker found display ads became far less effective at shifting purchase intent after the EU Privacy Directive restricted data-driven targeting - with the biggest losses on general-content sites and for small, static formats.

In the first industry-endorsed attention study, streaming video ads held almost 80% attentive viewing with minimal drop-off over time - 123% more attentive viewing than scrollable social (Amplified x Video Futures Collective, 2025).

Platform-run studies use marketer-invisible control construction and frequently disagree with independent geo holdouts, biasing toward the platform.

In a field experiment with an online travel firm, dynamic ads showing the exact product browsed were on average LESS effective than generic brand ads; specificity only wins once browsing behavior shows preferences have narrowed (e.g. the shopper has started visiting review sites).

Continuous or near-continuous presence generally beats short heavy bursts for established brands, because buying occurs continuously across the category.

Randomized field experiments on a restaurant-search platform show ads also lift sales for non-advertised competitors; spillovers concentrate on same-cuisine, highly rated rivals and are largest when ad intensity is low.

In two-minute windows around 1,224 commercials across 20 brands and $3.4B in spend, TV ads produce immediate, measurable lifts in site traffic and transactions.

A structural estimation approach recovers the complete MMM parameter set - adstock decay and saturation - directly from geo-experiments, giving the model causal calibration instead of observational guesswork.

Broad blocklists defund quality journalism inventory and shrink reach with little measured brand-safety benefit; contextual controls perform better.

Bing field experiments across thousands of brands show brand ads add a small but real 1-4% click lift when no competitor bids, and become defensive when rivals bid to siphon the traffic.

Google Research lays out how lift-test results should enter a Bayesian MMM as calibrated priors, with an explicit method for choosing which priors to tune; experiments become the model's anchor, not a parallel scorecard.

Outbound links in the post body still suppress reach. Algorithm InSights (1.8M posts) puts the body-link penalty near 50% of organic reach, and controlled A/B tests show first-comment placement recovering most of it: GrowthRocks measured 2.9x reach for link-in-comment in round one and 1.8x in round two — with more clicks, not fewer.

LinkedIn's own ranking model uses a dwell-time prediction weighted independently of — and more heavily than — likes, with no fixed second threshold. Content that gets read all the way through outranks content that gets tapped and skipped.

The 15x multiplier has no traceable primary source. AuthoredUp's analysis of 621,833 posts puts a comment at roughly 2x a like, a save at roughly 2x a meaningful comment, and threaded replies at up to 2.4x more reach. Comments still beat likes — just not by an order of magnitude.

Algorithm InSights (1.8M posts) documents organic reach falling by nearly 50% as LinkedIn shifted from a social graph to an interest graph. Baseline reach is structurally lower; the model now rewards demonstrated expertise signals over reach-chasing volume.

It replicates across networks. Sprout Social's LinkedIn test recorded 8,136 impressions for link-in-comment posts vs 3,309 with the link in the body (261 vs 141 engagements). Socialinsider's analysis of 51,054,216 Facebook posts found body links roughly halve engagement potential, and Hootsuite's linkless-tweet experiment found 56% of its top-engaging tweets contained no link.

Vector-autoregression modelling of a social network's growth shows WOM referrals have substantially longer carryover effects and substantially higher response elasticities than traditional marketing actions - and a referral's value can be monetized from the downstream ad revenue a new member generates.

Meta-analysis of 51 studies (339 volume and 271 valence elasticities): eWOM volume elasticity is 0.236 and valence elasticity 0.417 - with stronger effects on independent review sites and for durable, privately consumed, low-trialability products.

A 1.4-million-user randomized experiment on Facebook shows influence and susceptibility are separate, measurable traits - jointly targeting influential AND susceptible users outperforms chasing 'influentials' alone, and influence is demographic- and context-specific rather than a general property of people.

Analyzing every New York Times article emailed over three months plus follow-up experiments: positive content is shared more than negative, but the deeper driver is physiological arousal - awe, anger and anxiety spread, sadness suppresses - with surprise, interest and practical usefulness adding independent lift.

Analysis of branded video sharing shows high-arousal emotional content is what gets shared, yet sharing rates stay low even for strong creative - so earned views are a bonus on top of planned reach, not a channel.

Post-quality classification flags explicit engagement bait ("Comment YES if you agree") for downranking, and heavy traffic-bridging behaviour attracts minor reach caps. The tactic can still work for lead capture — it converts a smaller distribution into qualified DMs.

Socialinsider's benchmark data puts the average LinkedIn engagement rate at 3.85%, up 44% year over year, calculated per impression rather than per follower. Anything at or above ~4% is a defensible target.

AuthoredUp found no measurable reach impact from hashtags across an eight-month study. Hashtags are now a topical/discovery affordance, not a distribution lever — the interest-graph model classifies the post text itself.