How to Choose a PR and SEO Agency for AI Search: A Buyer's Guide

TL;DR

• The median enterprise B2B brand appears in 3% of the Google AI Overviews it's relevant for, across 45 million keywords (Walker Sands). In the other 97%, the answer was assembled from somebody else's pages.
• Of the 25 million links AI engines cited in Muck Rack's May 2026 study, 84% were earned: coverage, community threads, reference pages. Journalism alone was 27%. Paid content was 0.3%. Buying placements won't get you cited, and neither will a website nobody independent has vouched for.
• PR and SEO fail separately because each produces half a signal: a placement with no crawlable page behind it, or a thorough page with no corroboration.
• Before you hire anyone, ask who owns five jobs by name: audit, messaging, tracking, strategy, execution. Then score them with the card below. Score us with it too.

What is integrated PR and SEO for AI search?

Integrated PR and SEO means running earned media, your own site and the communities your buyers use as one program with one measurement, instead of three vendors with three dashboards. What you're after is being cited when a buyer asks ChatGPT, Perplexity, Gemini or Google's AI Mode who to consider.

Two terms you'll meet on the way. Generative engine optimization (GEO) is the work of making your brand's content and presence findable and citable by AI systems. Answer engine optimization (AEO) is the narrower craft of structuring a page so an answer engine can lift a direct answer from it. Both sit inside integrated PR and SEO. Neither replaces it.

Why integration matters comes down to where the engines look. Muck Rack analysed more than 25 million links cited in AI answers and found that 84% came from earned sources, the ones a brand neither owns nor pays for. Journalism was 27% of the total. Paid and advertorial content was 0.3%. Semrush's July 2025 study of more than 150,000 AI citations found Reddit dominated the most-cited domains on AI Mode, AI Overviews, ChatGPT and Perplexity alike. So the engines lean on coverage, community threads and reference pages far more than on your website. But your website is where they go to confirm the detail, and it has to be readable when they arrive.

An AI citation is also a different thing from a ranking. A ranking is a position for a query. A citation is a yes-or-no event inside one generated answer, and the same prompt can return different sources on different days. That difference is why measurement gets its own section below. If you want the longer explanation of how the engines choose, we wrote it in What is AI Visibility?

Why do PR and SEO fail when they're run separately?

Because each one produces half a signal and reports it as a whole.

Separate goals, separate dashboards

PR reports placements. SEO reports rankings and sessions. Neither report connects to the other, and neither can tell you whether an AI answer cited you. Corey Morris, writing in Search Engine Journal, lists five internal silos that quietly waste search budgets: strategy treated as a late tactic, channels that don't share research, content planned in isolation, data nobody can ladder up to revenue, and a web-development queue that never clears. We see at least three of these in most companies we talk to.

The placement with nowhere to go

A journalist quotes your founder. The article links to your homepage, or to nothing. The page that would have answered the buyer's next question was never written, because the SEO team didn't know the story was running. The engine records a mention with no destination.

The page nobody vouches for

The reverse failure. The SEO team publishes a thorough guide. No one independent has referenced it, so when an engine weighs sources, your page is the only one making the claim. Earned coverage is corroboration. Without it, owned content hits a ceiling.

Nobody can ship the fix

Morris tells of a restaurant chain where a change estimated at three days of development waited six months, because the IT team had no capacity and no way to let an outside agency in. We see the same thing with schema, redirects and crawler access. An audit that produces a PDF instead of a ticket with an owner has not produced anything.

How does PR Plus put them back together?

Althea Labs is a brand presence and AI visibility agency founded by Kara Silverman. We build brand presence and AI visibility programs for B2B startups through integrated PR, SEO and content, and PR Plus is how we do it. It has four components and runs in five steps. Here's the shape, with what "done" looks like at each step, so you can hold us, or anyone, to it.

The four components:

• Authority: traditional PR. Earned placements that signal to humans and machines that you're a legitimate option.
• Infrastructure: SEO and schema. The back end that makes your site readable to ChatGPT, Claude and the crawlers behind Google and Bing. Part 2 of the Althea Score is the checklist.
• Human proof: community. Genuine participation where your customers already talk. Reddit dominated the domains Semrush found cited across all four engines it studied, and in our experience a brand with no presence in those threads has no third voice to corroborate it. Part 4 of the Althea Score covers it.
• The bridge: content. Journalist-crafted posts, FAQs and pages that connect your PR authority to your community presence and answer buyer questions directly.

The five steps, and the acceptance test for each:

1. Audit. We review SEO content, technical SEO, PR footprint and AI visibility across the platforms that matter for your category. Accept it when every recommendation in the Brand Presence Report names an owner, a target and how you'll know it's done, whatever format it arrives in.
2. ICP and messaging. An ideal customer profile with buying triggers and research behaviour, a messaging framework, a channel map. Accept it when your sales team reads the buyer questions and says "yes, that's what they ask."
3. Visibility tracking. We set up Gauge to track 300 to 500 prompts pulled from real conversations and your own customer interactions, and report a baseline for AI visibility and SEO. Accept it when you can log in and see the baseline yourself.
4. Strategy. A 90-day editorial calendar across content, PR and community, every item tied to a visibility goal and a company goal. Accept it when you can point at any line and ask "why is this here?" and get an answer.
5. Execution. Writing, PR pitching, social and community, paid opportunities, speaking submissions and award nominations, plus a monthly report against the baseline. Accept it when that report separates placements, citations, visits and pipeline, and you have access to everything behind it.

What the first 90 days look like

Days 1 to 30 are the audit, the interviews and the baseline. You supply access to analytics, Search Console, the CMS and the CRM, and put your subject-matter experts on the calendar. Days 31 to 60 are the first outreach, the first pages and the first technical tickets. You return approvals inside the agreed window and assign a developer to accepted tickets. Days 61 to 90 are the first coordinated campaigns and the first report you can check against the baseline. Early movement typically shows in that window. The compounding is measured in months.

Which technical work actually makes a page eligible for an AI citation?

Eligibility is documented. Selection is not. Here is what the engines themselves say, which is the list we hold ourselves to.

• Google. There are "no additional requirements" to appear in AI Overviews or AI Mode. A page has to be indexed and eligible to show with a snippet, which means meeting the ordinary technical requirements: Googlebot isn't blocked, the page returns 200, the content is indexable. Structured data helps Google understand a page, and Google's own policies say it does not guarantee any feature.
• Bing and Copilot. Bing's guidelines say Search, Copilot and grounding results run on the same crawling, indexing and ranking foundation. Two directives matter more than most agencies mention: NOARCHIVE keeps a page's content out of Copilot responses entirely, and NOCACHE limits Copilot to your URL, title and snippet. Bing also asks for clear, consistent entity names, one topic per URL, and IndexNow notifications when pages change.
• ChatGPT. OpenAI says that for content to appear in ChatGPT summaries and snippets you must not block OAI-SearchBot. That's a different crawler from GPTBot, which governs training. Blocking one is not blocking the other. Check both lines in your robots.txt today; it takes a minute.
• Schema. Article markup describes headline, author, publisher and dates in machine-readable form. The test is simple: the markup names the same people and dates the reader sees.

Then there's the list we treat as experiments rather than requirements: FAQ schema, llms.txt files, fixed word counts, "short paragraphs," tables for their own sake. Some may help parsing on some engines. None is documented as earning a citation. An agency can run them, and we do, but each one belongs in an experiment register with a hypothesis, a baseline, a window, a cost and a stopping rule, and gets reported as an experiment. Anyone who tells you schema guarantees a citation, or that one format satisfies every engine, has just told you what they don't know.

How do you measure AI citations through to pipeline?

Keep four records, and don't let a report blend them.

1. The placement. Campaign, angle, outlet, URL, date, spokesperson, destination page, and whether the link was crawlable.
2. The citation. A fixed prompt set, run on a schedule, logging the engine, the date, whether you were mentioned, and the exact URL cited. Run each prompt three to five times per check, because the same prompt returns different sources on different runs, and report the run count. Treat small movements as noise.
3. The visit. ChatGPT adds utm_source=chatgpt.com to referral links, so clicked ChatGPT citations show up in GA4 acquisition reports. Google is the opposite: clicks from AI Overviews and AI Mode are folded into ordinary Search traffic in Search Console with no breakout. Perplexity and Gemini appear as referrers only when someone clicks. Nothing measures a citation that was read and not clicked.
4. The outcome. GA4 key events for demo requests and form fills, passed to the CRM as leads and opportunities. Keep user, session and event acquisition apart; GA4's traffic-source dimensions have different scopes, and mixing them double-counts.

Then label every conclusion by the evidence under it:

• Observed: the records show a placement, citation, visit or lead happened.
• Associated: citations or outcomes changed after a campaign; the records don't say why.
• Attributed: an agreed analytics or CRM rule assigned the credit.
• Causal: a controlled design isolated the campaign's effect. Rare. Say so when you have it.

No agency can prove which placement made an engine cite you. The engines draw on several sources, change their retrieval, and answer differently on different days. What an agency can show is sequence and association, honestly labelled, and the two rates that matter: citation rate (answers citing your domain divided by answers sampled) and lead rate (qualified leads divided by referral sessions). A report that turns every organic lead into attributed revenue with no agreed rule is the loudest red flag in this guide.

How should you score a PR and SEO agency?

Score each category from 0 to 5, multiply by its points, and divide by 5. A 4 on team integration is 4 × 20 ÷ 5 = 16. The points add up to 100, so the maximum total is 100.

The scale: 0, no evidence. 1, a verbal assurance. 2, a generic document that could be any client's. 3, partial evidence specific to you. 4, complete evidence with one gap. 5, complete evidence you could check yourself.

Our suggestion for the cut: shortlist at 70 or above with no category below 3. Under 60, don't shortlist, whatever the pitch.

These categories reflect how we structure PR Plus. Re-weight them before you score anyone, including us.

Four things disqualify an agency whatever its total. This is our position, not a citation.

• An unnamed delivery team. If you can't meet the people before you sign, you're buying a brand, not a team.
• Keyword-export-first discovery. An agency that opens with a keyword list hasn't asked how you make money or how your buyers decide.
• Guaranteed citations. Nobody controls what an engine includes. Google's own advice on hiring an SEO says the same about rankings: if they guarantee first place, find someone else.
• Vague ownership language. If the contract doesn't say the accounts, the data and the files are yours, assume they aren't.

What does it cost, and should you build it in-house?

What it costs

Ask for the whole number. Our engagements start with a Brand Intelligence Audit, a one-time project delivered in five days. From there the work runs as a fixed-scope toolkit, a three-to-six-month sprint, or an ongoing monthly program; the packages are on the services page, and we'll quote them against your scope. Whatever you're quoted, by us or anyone else, ask what's not in the number: content production, technical implementation, tools, reporting software, revisions. A low retainer with production billed later is a high retainer.

In-house or agency?

In-house works when you have three people: a writer with a journalist's instinct for what will land, someone who can ship technical changes to the site within days, and someone who owns measurement and will re-run the prompts every week without being asked. The Series A and B teams we meet usually have one marketing lead who is all three on paper and none of them in practice. That's when an agency makes sense, and the honest version is a hybrid: the agency owns strategy, earned media, community and measurement; your developer ships the tickets; your founder shows up for interviews. If you already have the three, buy the audit on its own and run the rest yourself. We sell it standalone for exactly that reason.

How is an AI-visibility agency different from a traditional PR firm?

A PR firm's product is the placement. Ours is presence across every source an engine draws on when it answers a buyer: earned coverage, your own site, the communities where your customers talk, and the reference pages that describe your category. The placement is one input. Before we pitch anything we make sure the site is technically eligible to be cited, because a story that lands on a page an engine can't read is a story we wasted. We baseline 300 to 500 real buyer prompts and re-run a tracked set every month, so what we report back is citation rate rather than a coverage list. And community is a channel on our calendar, not a risk we monitor. A traditional firm can tell you whether you were covered. We can tell you whether you were cited, and whether anyone came.

FAQs

How long does it take to see results? Your baseline lands in the first month, alongside the audit, so you have a number to argue with early. In our own engagements, technical fixes, first placements and community presence have shown measurable movement in the 60-to-90-day window; the compounding takes longer than that. Treat any agency's timeline, including ours, as a claim to check against your own baseline rather than a promise.

Does PR replace link building? No. PR earns links and mentions as a by-product of coverage; link building targets links directly. Google treats crawlable links as a discovery signal, not the goal. Expect an agency to coordinate both without reducing every story to its backlink value.

Who owns technical implementation? The contract should name an owner for every recommendation. Usually the agency audits, writes the ticket and validates the result, and your developer deploys it. Agree access, approval and deployment responsibility before signing, or the tickets sit in a queue for six months.

Can an agency guarantee citations? No. Engines choose sources independently and answer differently by platform, wording, location and date. An agency can make you eligible, track the prompts that matter, and report what it observes. That's the whole promise, and it's enough.

Score us too

Run the card above on us. Every Althea engagement starts with an audit: where you stand in search, in AI answers, and in the communities where your buyers already are, with a baseline you can log in and see. Start with an audit.