880 people a month search "answer engine optimization services" (Google Keyword Planner, All locations, 2026-08-20), up 48% in three months and 306% year over year — the fastest-growing of the AI-search terms we track. Four advertisers are already bidding on it. Before you hire anyone, here's what we found when we tested the market ourselves: who Perplexity actually recommends, what the pages Google's own AI cites have in common, and what happened when we ran this exact play on a real manufacturing client.
What you're actually buying isn't one thing
We asked Perplexity, phrased the way a buyer would ask it: "who can I hire to get my manufacturing company recommended by AI chatbots like ChatGPT." Its answer — drawn from 29 sources — didn't name a single vendor. It split the work into four separate roles:
| Hire | Best when | What they should own |
|---|---|---|
| Full-program AEO agency | You need the whole thing done | Technical SEO, industrial content, AI-search measurement, digital PR |
| Independent technical SEO consultant | Your site has crawl, speed, or CMS problems | Indexation, robots.txt, structured data validation |
| Industrial content expert | Your expertise is stuck in engineers' heads | Application guides, spec-comparison pages, case studies, test data |
| B2B PR / trade-media specialist | You need third-party credibility | Trade coverage, expert quotes, association listings, real citations |
That's a useful filter even if you never ask Perplexity yourself: a vendor who sells you one bundle for every problem is selling convenience, not a diagnosis. Ask which of the four boxes your site actually needs before you sign anything.
Not sure which of the four your site needs? Run the free check — three results on the spot, no account, no sales call.
The one promise no honest vendor will make
Perplexity's own answer to "who should I hire" said it directly: "no firm can guarantee that ChatGPT will recommend you for a given prompt." If a sales page promises guaranteed AI recommendations, that's the tell to walk away.
The honest version of the job is narrower and more useful: make your business easy to discover, verify, cite, and contact when an AI system answers a question your buyer actually asked. That's a testable standard. "Getting recommended" isn't.
We put the same question to ChatGPT — logged out, no account attached, nothing for it to draw on but the question itself. Its answer landed on the same line, unprompted: "AEO isn't a scam. 'Guaranteed ChatGPT rankings' is the scammy part." Three separate systems, asked separately, gave the same read.
What actually worked: asking the AI, then checking the answer
Most advice about AI visibility is theoretical. We ran the test on a real client instead.
What we asked, and what came back
We asked Perplexity, twice, in different wording, to recommend suppliers in one manufacturing client's exact product category. It named that client first both times, citing pages from their own site — full detail in our manufacturer AEO case study.
The scale check that keeps this honest
The win matters less without the honest number next to it: over 90 days that client's analytics show five sessions from chatgpt.com out of roughly 417 total. AI referral there is real and mechanically verifiable — and currently small.
That result matches exactly what Perplexity itself recommends buyers ask for: "a recurring test set of real buyer prompts across ChatGPT, Google AI features, Perplexity, and standard Google search, tied to qualified visits and pipeline — not just chatbot mentions." If a vendor can't show you a test like that on a real client, ask why not.
Check the one thing that silently blocks all of this
Before content or citations matter at all, confirm your site isn't accidentally invisible to the crawler that feeds ChatGPT Search. OpenAI's own documentation names it OAI-SearchBot, and says allowing it is a separate setting from allowing GPTBot (which is used for model training, not search). A minimal robots.txt line that allows it looks like:
User-agent: OAI-SearchBot
Allow: /
Changes to robots.txt take roughly 24 hours to be reflected, per OpenAI's guidance. It costs nothing to check and takes under five minutes — do this before paying anyone for anything else.
We checked our own robots.txt too
Not just OAI-SearchBot — Anthropic's ClaudeBot and Claude-User, Perplexity's PerplexityBot and Perplexity-User, and Common Crawl's CCBot are all named and allowed on our own site, right now. It's a two-minute check on any site: pull /robots.txt and look for a Disallow sitting under any of those names. If one's missing or blocked, that's the first thing to fix — before a single word of content.
What Google's own AI actually rewards here — we watched it happen
This is the part most "AEO services" pitches skip: we pulled Google's actual AI Overview for this exact search term and checked which pages it cited, then opened those pages and counted what was really on them.
Ranking and being cited are two different games
One page cited twice by Google's AI Overview doesn't appear anywhere in the top 30 natural results — while a page ranking second was never cited by the AI at all. Authority didn't decide who got quoted.
Three pages, side by side
Here's what we found comparing three of the cited pages:
| Cited twice, ranks #1 | Cited twice, not in top 30 | Ranks #2, never cited | |
|---|---|---|---|
| Body word count | 3,442 | 1,964 | 1,119 |
| H2 + H3 headings | 23 | 52 | 5 |
| JSON-LD schema | 0 | 1 (bare WebPage only) | 0 |
Why the most-headed page won
The page with the most headings — not the most words, not the most schema — got cited the most. And a direct phrase match confirmed why: Google's AI Overview breaks this topic into five components, and one of them names "knowledge graph alignment (Wikidata, entity mapping)" — a phrase that appears, word for word, on exactly one of the three pages we checked. The AI Overview's structure wasn't generated from nothing. It was assembled from the subheadings of the page that used the most specific, most quotable language.
The takeaway you can act on today: structure your page around the specific questions a buyer would ask as separate subheadings, and use precise, quotable terms instead of generic ones. That's a content decision, not a technical one — and it's free. Check your own site's AI-readiness free if you want to see where yours stands.
AEO and GEO — same idea, two labels
"Answer engine optimization" and "generative engine optimization" are close enough in practice that most agencies use them interchangeably. If you want the fuller picture — including why the market sells five near-identical names for the same underlying work, and what Google's own documentation says three of the commonly-sold tactics don't actually require — we wrote that up separately in our GEO explainer.
What this costs, in public
Advertisers are currently bidding MYR 5.44 to MYR 103.95 per click on this exact search term (Google Keyword Planner, 2026-08-20). Four paid ads are running on the first page right now, including one — One IMS — advertising specifically to manufacturers. That range is wide because it's blending very different scopes: a single technical audit costs far less than a full monthly AI-visibility program. Get a quote scoped against one of the four roles in the table above, not a flat number.
Where to go from here
If you want to check this yourself first
Run the robots.txt check above, then ask Perplexity or ChatGPT directly who they'd recommend in your category — that single test tells you more about your current visibility than any sales pitch will.
If you want it done for you
Run the free check — three results on the spot, no account, no sales call.
If you're comparing what things should cost
See our breakdown of what SEO and AEO work actually costs, with sources.