What a web SEO agency should actually deliver

Ask a marketing leader what they hired last year and you will hear “SEO.” Ask what actually shipped and you will hear a pile of keyword decks, a homepage that still takes 4.1 seconds to become interactive, and a blog that reads like every other SaaS competitor in the SERP. A web SEO agency is not a content mill with a Search Console login. It is the team that treats the website as the ranking product: templates, crawl paths, Core Web Vitals, offer pages, and the kind of information gain Google’s 2026 ranking stack actually rewards.

That distinction matters because the click is no longer guaranteed. Google reported AI Overviews reaching over 1 billion users monthly by late 2024, with rollout continuing through 2025-2026. Zero-click searches now account for an estimated 60-65% of US queries (Sparktoro, 2024), a share that has grown with each AI Overview expansion. When an Overview sits on more than 20% of queries, organic CTR can fall by nearly 60% even as impressions look “healthy.” You do not need another dashboard of keywords. You need a partner who can make the site worth citing, worth clicking, and worth converting after the click.

This piece is the operator’s view of that work: where web SEO programs stall, how a serious agency actually runs the site, a scorecard you can use on your own properties, and the measurement that still holds when last-click attribution is a joke.

When “we do SEO” still leaves the website unrankable

Plenty of retainers start with a technical audit PDF and never touch the CMS. The agency flags 200 “issues.” Your developers fix robots.txt. Six months later the money pages still share one thin template, the H1 on the pricing page is the brand slogan, and INP on mobile is a mess. May 2026 CrUX data showed only 55.9% of origins passing all three Core Web Vitals. If you are in the failing half, Google does not care how clever your cluster map looks.

Buyers already speak this language. They ask “who specifically will manage my account?” They say “if they send traffic to your homepage, run.” They have watched 12.5% of SEO relationships disappoint on outcomes and another slice of reviews complain about outsourced content that arrives late and generic. Generic AI copy now saturates the index; differentiation comes from information only your team can provide. Volume is a dead metric. Information gain is what moves the needle after the March 2026 core update.

The failure mode is operational, not mystical. You publish 8 posts a month that restate the same three talking points. You never rebuild the comparison page that actually sits on commercial intent. You never map schema to the FAQ your sales team already answers on calls. You never check whether the page that ranks is the page that books a demo. A web SEO agency that cannot own templates, internal links, and conversion paths is doing research theater.

Search CPCs have climbed year over year across most B2B categories, making it harder to keep buying the same query while organic sits on page two. Paid search still prints pipeline, but you cannot keep buying the same query while organic sits on page two of an AI-heavy SERP. The site has to earn citation share and the remaining clicks. That is website work, not a monthly keyword spreadsheet.

If organic sessions look “fine” while demo quality is sliding, the leak is usually the page, not the algorithm. HeyLead SEO services start from that page, not from a vanity ranking slide.

What a web SEO agency actually changes on the site

The useful unit of work is a money URL, not a “content calendar.” You pick the pages that already attract sales conversations (pricing, integrations, industry use cases, competitor alternatives) and you make them faster, clearer, and denser with first-hand proof than the 10 URLs Google already trusts.

Technical work is the floor. Crawl budget wasted on faceted junk, canonical chains, orphaned case studies, and bloated JavaScript that wrecks INP will bury even a strong brief. Then comes information architecture: one primary intent per URL, supporting clusters that actually link, and breadcrumb plus FAQ schema that answer-engines can quote. Then comes the human layer Google still cannot fake well: named practitioners, original screenshots, pricing logic, implementation caveats, and the ugly trade-offs your category pretends do not exist.

AEO and GEO sit on top of that, not beside it. Being the brand mentioned in AI results means E-E-A-T, citation-worthy specifics, and SERP impression visibility even when CTR is ugly. Search Console’s generative AI reporting is now part of the weekly read, not a novelty tab. You watch which passages get pulled, then you strengthen those passages instead of spinning another 1,800-word explainer that says nothing new.

Post-click is part of SEO. If the ranking URL dumps a high-intent visitor onto a 14-field form with no social proof, you did not “win SEO.” You paid for an impression that bounced. Session behaviour (scroll depth, rage clicks on the nav, form abandon) tells you whether the SEO brief or the layout is the bottleneck. That is why HeyLead Insights belongs in a web SEO program: you watch the same URLs you are trying to rank and you fix the friction that kills qualified leads.

Cadence still matters, but not as a vanity publish count. A small set of pages that compound beats a firehose of AI drafts. When you need volume without a 12-person writing team, a managed system like HeyLead Auto Blogger can ship niche-aware drafts under a human strategy. The editing pass is non-negotiable. Ranking is not “easier with AI.” It is easier to flood the index with mush, and Google is not confused about the difference.

A working scorecard for your money pages

Use this on 8-12 URLs that already show commercial intent in Search Console. Score the page, not the agency pitch. If more than half the list fails, you do not have an SEO problem in the abstract. You have a website that cannot carry the channel.

Two site-level bottlenecks that looked like “the algorithm”

A VP of Marketing at a mid-market B2B software company inherited an SEO retainer that published 12 posts a month. Impressions climbed. Demo requests from organic did not. The commercial cluster all pointed at a homepage hero that rotated four unrelated products. Search Console showed the highest-impression query landing on a 2019 “what is” article with no CTA. They collapsed four product H1s into one services template, moved proof above the fold, and redirected the explainer to a comparison URL sales already used. Organic-assisted opportunities moved within a quarter. The algorithm had not “turned on.” The site had finally matched the query.

A founder-led professional services firm ranked for brand plus a handful of informational terms. They wanted “web SEO agency” type commercial queries without touching templates because “content should be enough.” Mobile INP on the contact page failed because a scheduling embed loaded four unused scripts. Desktop conversion looked acceptable. Mobile, where most new sessions arrived, lagged. They stripped the embed to a single lightweight form, cut fields from 11 to 5, and added a 90-word scope note that mirrored the sales call. Rankings barely moved. Qualified form volume did. That is still SEO: the ranking URL has to complete the job.

Neither story required a new channel. Both required someone who could change the website, not just annotate it. Realistic organic programs still take 3-6 months to show compounding results. Anyone promising a GEO miracle in a week is selling a slide, not a site.

When paid is already in the mix, keep the same landing thesis. Do not run Performance Max to a homepage while organic fights for a dedicated URL. Align query, ad, and page. 20-30% of SEM waste still comes from irrelevant queries and misaligned landings. Organic inherits the same tax if the page is wrong.

Two patterns we hear repeatedly from marketing leaders taking over inherited retainers: impression growth celebrated while the comparison page that actually created opportunities still had a slogan H1 and a 4-second INP - treating that URL as the product is when pipeline from organic stopped being a mystery. And the first question that exposes a hollow retainer is still “who will touch the CMS?” Silence is the whole answer.

Measurement that survives AI Overviews and messy attribution

You will hear “attribution is basically dead.” Last-click is. Channel scorekeeping still helps you decide which URLs to fix this month. Track ranking URL, impression share in classic and AI surfaces, click-through on the remaining blue links, on-page conversion rate, and downstream opportunity creation. Do not let a pretty average position hide a 60% CTR collapse under Overviews.

Clean conversion setup still matters for any paid support around organic (Enhanced Conversions, server-side events) because 20%+ of conversion data still disappears into ad blockers and privacy gaps. For SEO itself, first-party events on the site are the ground truth: form start, form success, calendar book, qualified flag from CRM. If marketing and sales disagree on “lead quality,” fix the definition on the thank-you event before you debate the content brief.

Reporting cadence should be weekly enough to catch a template regression, not a 40-page monthly PDF that buries the INP failure. Name the person who owns the account. Vague retainers at $3,000 median, or full-stack fees that balloon without a named lead, are how programs stall. The most common complaint in agency reviews isn’t rankings - it’s not knowing who’s actually doing the work.

Worth paying a web SEO agency versus running Ahrefs in-house? In-house tools find gaps. They do not rebuild templates, negotiate with engineering, write information-gain pages, or sit in the CRM when “organic leads” turn out to be students. The fee is for that operating load, not for a login.

FAQs

How is a web SEO agency different from a content shop?

A content shop ships articles. A web SEO agency changes the URLs that can rank and convert: templates, internals, CWV, schema, and the commercial copy. If they cannot ship a production change, they are a research vendor.

How long before commercial queries move?

Plan on 3-6 months for compounding results on competitive terms. Technical and conversion fixes on already-indexed money pages can show lead-quality gains faster. One-week GEO promises are not a C-level plan.

Should we still invest if AI Overviews steal clicks?

Yes. Track your Overview impression share separately in Search Console’s AI mode filter. If you appear in Overviews but CTR has fallen, the fix is strengthening the passage Google is already quoting - not writing a new page. Visibility without a click can still influence a later branded search. Thin pages that only existed to harvest clicks are the ones that die first.

What should we keep in-house?

Offer, proof, and subject-matter review. Let the agency own crawl health, information architecture, on-page systems, and the publishing machine. You still approve anything that hits the market with your name on it.

Do we need paid search if organic is the goal?

Often yes on the same commercial clusters while pages mature. Paid also reveals which messages convert, which you then put on the organic URL. Do not treat the two as rival scoreboards.

Putting it to work

Pull the 15 organic landing pages with the most impressions in the last 90 days and score them against the checklist above. You will usually find two commercial URLs that deserve a rebuild before you write another blog post.

This week

• Export Search Console landing pages and mark commercial vs. educational.

• Fix H1 and title on the top 3 commercial URLs.

• Run Core Web Vitals on those URLs and log INP offenders.

• Add 5 internal links from high-impression posts into one money page.

• Validate schema and indexation on production.

• Watch form starts vs. completions on those URLs for seven days.

Next 30 days

• Rebuild one comparison or pricing template with unique proof and a shorter form.

• Align one paid cluster to the same URL organic is targeting.

• Stand up weekly reporting on URL-level qualified leads, not average position.

• Decide which pages a human must write vs. which can draft through a managed cadence.

If the bottleneck is the website itself (templates, CWV, information gain, and the conversion path on ranking URLs) and you do not want that load sitting on a stretched in-house team, HeyLead will run the web SEO program end to end. Chat with us on WhatsApp