The modern PPC guide —
written for AI-run auctions.

01Why paid media still matters — even as AI rewrites the auction

Paid media gets you qualified traffic on day one. Unlike organic, there's no ramp — budget goes in, traffic comes out, the feedback loop closes within hours. That's why every growth company funds it even when they're compounding organic in parallel.

What's changed is what you actually control. Five years ago, an experienced paid-media manager adjusted keyword bids, audience combinations, and dayparting schedules daily. Today Google's smart-bidding systems, Meta's Advantage+, and LinkedIn's Predictive Audiences run those decisions at millisecond cadence against signals you can't access manually. Manual bid tuning underperforms in most accounts now.

The old skill was bidding. The new skill is feeding the bidder the right signals, the right creative, and the right conversion values — and knowing when to override it.

The chart below shows blended CPL by platform from our 2025 engagement dataset (B2B, SaaS / professional services). Your numbers will differ by vertical and offer, but the shape is consistent: Google captures intent at medium cost, LinkedIn costs 3× but the leads are precise, TikTok and Meta deliver volume at low cost, emerging surfaces are a wildcard.

02The four-platform landscape: Google, Meta, LinkedIn, and the rest

Each platform rewards a different intent layer. Get this right and your mix works. Get it wrong and you're buying traffic in the wrong place.

Google (Search + Shopping + PMax)

Captures existing demand. Someone types a query — you bid for the click. Best for products/services with clear commercial intent searches. Search stays human-friendly; Performance Max is closer to an AI black box that expects clean conversion signals and refuses to work with them.

Meta (Facebook + Instagram)

Generates demand. People aren't searching — you interrupt their scroll with creative good enough to earn attention. Best for products with visual appeal, strong offers, or broad relevance. The Advantage+ shopping campaigns now handle audience expansion automatically; creative is 90% of the performance.

LinkedIn (B2B)

Unmatched targeting at company + role + seniority level. Expensive (CPLs north of $100 are normal) but the lead quality is unmatched for enterprise B2B. Essential for deal sizes over $15K, optional below.

Everyone else: TikTok, YouTube, X, Reddit, Pinterest

Run as experiments, not core channels. Each has a distinct audience + ad format — TikTok for trend-led commerce, YouTube for consideration and branding, Reddit for intent-rich niches, Pinterest for visual discovery. Allocate 5–10% of budget here and test seriously before promoting a winner.

03Landing pages are half the campaign

The biggest single lever on paid CPL isn't in the ads platform — it's on your landing page. A 2× landing-page conversion rate halves your CPL across every channel, every keyword, every audience.

What actually moves the needle

• Intent match. The headline above the fold must mirror what the searcher asked. “Buy wireless earbuds” should land on a page headed “Wireless earbuds” — not on a homepage carousel.
• Form design. Short forms convert better, but “email only” attracts garbage. A 3-field form (name, email, company) is usually the right trade-off. Multi-step forms with progress indicators beat long single-page forms for higher-consideration offers.
• Load time. Every 1s of additional LCP costs ~10–15% of mobile conversions. Pre-rendering, image optimisation, and minimal JS on landing pages are non-negotiable.
• Trust signals. Real logos of real clients, real testimonials with real names, real numbers. Stock reviews and fake urgency tank conversion — people aren't stupid.
• One clear CTA. Every landing page has a single action. If you need to offer both “start trial” and “book demo”, split into two pages and let the ad match.

We cover landing-page architecture in more depth on the PPC landing-pages service page, including the AI-driven variant testing we run.

04Bid strategies, smart bidding, and when to override the algorithm

The quick reference:

• Target CPA — tell Google/Meta what a conversion should cost. Good once you have 30+ conversions/week.
• Target ROAS — feed conversion VALUES (not just counts) and optimise for revenue. Essential for e-commerce.
• Maximise Conversions — let the algorithm spend your daily budget. Only use if you don't have target data yet.
• Maximise Conversion Value — same as above but value-weighted. Needs clean value signals.
• Manual CPC — useful for low-volume accounts, brand-defence campaigns, and early-stage testing where you need tight control.

The override cases: when you have CRM data showing certain lead types are worth 5× more, feed that back via enhanced conversions so the bidder can optimise for high-value leads (not just count). When the algorithm's “learning phase” drags on beyond 10 days, your signal is too weak — consolidate campaigns or raise budgets. When Target ROAS drifts over three weeks, check for audience overlap or creative saturation before blaming the bid.

05Creative: the only remaining lever at scale

Because the auction is largely automated, creative is the place human judgment still decides who wins. On Meta, the same audience + same offer + same budget will produce 3–5× different returns depending on whether the creative resonates. That's not a margin — that's the whole ball game.

The creative cadence that actually works

• 4–6 new concepts per month, minimum. Meta creative CTR halves within two weeks of launch (see Figure 03 above). Refresh cycles under 14 days keep the flywheel moving.
• Variant testing within each concept. Hook, opening frame, voiceover, and CTA — test systematically. 1 winner per 3–4 variants is a reasonable hit rate.
• AI-accelerated production. LLMs draft scripts to a brief; designers + video editors ship at 3× the old pace. The editorial hand stays human — it's the repetitive production work that's been compressed.
• Platform-native formats. Meta Reels ≠ TikTok ≠ YouTube Shorts. The content might be similar, but hooks, pacing, and subtitling differ enough that cross-posting alone loses 40–60% of performance.
• UGC and creator content. Perform significantly better than studio content in direct-response. Structured creator programs (specific briefs, revision cycles) beat ad-hoc influencer drops for reliability.

06The AI layer: where the real leverage lives now

“AI in paid” isn't smart bidding — that's been around since 2019. The real AI layer in 2026 is what you do with the conversion data, creative pipeline, and campaign structure sitting above the platforms.

What AI is doing for operators who use it well

• Enhanced conversions with LTV signals. Feed the platforms not just “lead converted” but “this lead is worth $X predicted LTV.” Smart bidding then optimises for revenue, not lead count. The gap between a count-only account and a value-informed account is huge.
• AI-generated variant creative. Large language models draft ad copy to a tight brief; diffusion models produce image variants; video models handle short-form variations. Human editing on top. Production tempo hits 4–6 net-new concepts per week for a single brand.
• Predictive audience signals. First-party data run through a model to predict which segments will convert, then uploaded as custom audiences. Beats platform lookalike modelling when your first-party signal is strong.
• Automated anomaly watch. Instead of a media buyer checking accounts daily, an AI watches pacing, CTR, CPL, and ROAS against learned baselines and flags the human only when something deviates. Frees the senior paid operator from dashboard-babysitting.
• Offline conversion upload. CRM-tracked events (SQL, closed-won, revenue) fed back to the platforms so the bidder optimises for real outcomes, not top-of-funnel proxies.

07Attribution — and why last-click is still breaking your budget

Paid-media platforms report on what they delivered. Each claims credit for the same converted customer. Add up all the platform claims and you'll find you sold 3× more product than actually left the warehouse.

The fix is a unified attribution view that reconciles:

• First-party data. Your CRM, your billing system, your product analytics — the single source of truth.
• Platform-reported conversions. What Google / Meta / LinkedIn say they drove.
• Cross-device + cross-domain signals. How visits become leads become customers across multiple touches.
• Modelled attribution. Google's Conversion Modelling and Meta's aggregated event measurement fill in the gaps that privacy changes opened (iOS 14+, cookie loss).

Practically: fractional multi-touch attribution at the CRM layer, run either through a specialist tool or through InsightAX where we own it. Every closed-won deal has a full trail including every paid touch — so you can see which campaigns actually influenced revenue, not just which got credited by the last platform they touched.

08Scaling without bleeding

Most paid accounts hit a ceiling at some budget level where CPL starts climbing fast. The naïve response — “we need 20% more budget to hit the number” — usually makes it worse. Three better moves:

• Diversify the audience/keyword base. If you're maxed on your top 5 ad groups, open the next 10 — broader-match terms with tight negatives, lookalike expansions, related product categories.
• Widen the creative funnel. Top-of-funnel prospecting creative plus mid-funnel consideration plus bottom-funnel retargeting. Each does different work. Most accounts over-invest in bottom-funnel and starve the top.
• Expand platforms. If you're 80% on Google, you're one algorithm update from a bad quarter. A balanced 50/30/20 (Google/Meta/LinkedIn or other) gives you resilience and lower blended CPL as each platform hits diminishing returns at different budget levels.

The chart below shows the typical ROAS trajectory for accounts that scale correctly — a learning-phase dip, then compounding as audiences stabilise, creative rotates, and attribution signals mature. Accounts that quit in month 2 are quitting right before the curve starts bending.

09Working with a paid-media partner

Questions worth asking any agency, consultant, or in-house candidate you're evaluating:

• Who is going to touch my account daily? If the answer is “our AM coordinates, and we have specialists,” keep looking. You want the named operator who will be in the account every day.
• How do you handle creative? A serious shop has a creative pipeline (in-house or partnered), not “we'll use whatever you send us.” Creative IS the lever — if they don't own it, they don't own the outcome.
• How do you measure contribution to revenue? Platform reports alone are a red flag. Insist on seeing how they reconcile with your CRM.
• What's your view on smart bidding? “We use it” is not an answer. A strong one explains when they override, what signals they feed it, and how they monitor for learning failures.
• What's your scaling methodology? If the answer is “we'll push budget when CPL looks good,” they'll blow your Q4.

That's how we run paid at Axccelerate — named operators on every account, creative pipeline owned in-house, enhanced conversions and CRM-tied attribution as standard, scaling playbook documented per client. If that's the shape of operator you're looking for, the next step is a 30-minute scoping call and a scoped proposal within two working days.