A growing share of the questions that used to go to Google now go to ChatGPT, Claude, Perplexity and Gemini. "Who is the right person to talk to about X?" "Is this company credible?" "Who is Jane Park?" The answer comes back as a paragraph with a handful of citations. You are either in that paragraph or you are not, and most professionals have no idea which.

Answer Engine Optimization, AEO, is the practice of making a person or a company resolvable, trustworthy and citable to those engines. It is not SEO with a new name. SEO wins you a position in a list of links. AEO wins you a place inside the answer. The mechanics are different, and the thing most people assume is their professional presence online, their LinkedIn profile, turns out to be invisible to the engines doing the answering.

This piece explains how the engines actually build an answer about you, what they weigh, and what you can do about it.

How an AI engine builds an answer about you

When someone asks a search-enabled engine about a person or a company, it does roughly the same five things every time.

  1. It retrieves. It runs live web searches (ChatGPT leans on Bing, Gemini on Google, Perplexity and Claude on their own or partner indexes) and fetches the top pages.
  2. It extracts. From each page it pulls facts: name, role, employer, topics, locations, dates. Structured data is read directly. Prose is parsed, and parsed prose can be misread.
  3. It resolves the entity. It decides whether the Jane Park on one page is the same person as the Jane Park on another, using matching names and employers, links between the pages, and explicit sameAs declarations.
  4. It corroborates. A fact stated by several independent pages outranks a fact stated once. Your own page counts, but it counts more when other pages agree with it.
  5. It composes and cites. It writes the answer and cites the pages it leaned on. A page it could not fetch, or could not parse, does not exist for this step.

Two things follow from this. The engine's picture of you is only as good as what it can crawl. And that picture is assembled from many pages, so the useful question is not "is my profile good?" but "do the pages about me agree with each other, link to each other, and can the engines read them?"

The inputs that matter

The engines do not publish their weights, and they differ from one another. But most of what matters follows directly from how retrieval works, and the rest shows up consistently when you run the same cold queries week after week and watch what changes. Here is the list.

Crawlability. Can the engine's bot fetch the page at all? That means a robots.txt that allows GPTBot, ClaudeBot, PerplexityBot, Google-Extended and the unnamed agents, no login wall, and plain HTML that loads fast. This is the gate everything else sits behind.

Machine-readable structure. Can it extract facts without guessing? Schema.org JSON-LD for Person and Organization, with jobTitle, worksFor, knowsAbout, sameAs and dateModified, removes the guessing.

A canonical page. Is there one URL the engine should treat as the source of truth? One page per person, one per company, on a domain the subject controls, with every duplicate redirecting to it.

Corroboration. Do independent pages agree? Your employer's site, your personal site, your social profiles, conference bios and press should all state the same role and link to the canonical page. Agreement across sources is what turns a claim into a fact.

The link graph. Do the pages about you point at each other? Person to company, company to person, both to outside profiles via sameAs, and those outside profiles pointing back. Links are how the engine knows two pages are about the same entity.

Freshness. Is this current? A visible "last updated", dateModified in the structured data, and recent additions. Stale pages lose to recent ones when they conflict.

Topic evidence. Why would it cite you on a subject? Long-form writing on your own domain, talks, named areas of expertise, each tied to you as an entity. An engine cites people it has evidence about, not people who describe themselves as experts.

Question-shaped content. Does a page answer the question the way it is asked? Plain statements of role and scope, FAQ blocks, and directory pages like "people at Acme who work on treasury payments" map directly onto the queries people type.

Agent-readable channels. Can an AI agent read you without scraping? llms.txt and, increasingly, a public MCP endpoint let agents pull the facts directly.

Notice what is not on the list: follower counts, engagement, posting frequency, how often you comment. The engines are not reading your feed. They are reading pages.

Why your LinkedIn profile is invisible to the engines

For most professionals, the richest record of their career is their LinkedIn profile. For the engines, it does not exist.

LinkedIn's robots.txt, as of October 1, 2026, blocks GPTBot, ClaudeBot, PerplexityBot and every unnamed crawler outright. It allows only the classic search-index bots, with restrictions, and blocks the agents that fetch pages live at answer time. Profiles sit behind a login wall for most visitors. The structured data on a profile is minimal and belongs to LinkedIn, not to you. Nothing you write there carries sameAs to your other presences, and nothing you publish there is on a domain you control.

This is a business decision, not an oversight. LinkedIn's model depends on being the place the data lives and charging for access to it. Letting AI engines read it freely would undercut that. So the walls will stay up.

The consequence is that when an engine is asked about you, it falls back to whatever else it can find: a conference bio from 2019, a quote in a trade article, a company team page nobody has updated since you were promoted. Fifteen years of career history, carefully maintained, contributes nothing. The engine answers from scraps.

The same is true for companies. A company page on LinkedIn, and the hundreds of employee profiles that point to it, is the strongest body of evidence that exists about what a company does and who does it. The engines cannot see any of it.

What to do about it

None of this requires a budget. It requires owning a page.

  1. Put one canonical, crawlable, structured page about yourself on a domain you control. A personal site works. It needs Person JSON-LD, a visible last-updated date, and a robots.txt that lets the engines in.
  2. Point everything else at it. LinkedIn's website field, your X bio, your email signature, your speaker bios, your company's team page. Each one is a vote that the page is you.
  3. Link back. Your page should carry sameAs links to every profile you own, and your company's page and yours should link to each other.
  4. Publish on your own domain. Articles you have written for LinkedIn are invisible there. Republish them where they can be read, with you as the author entity.
  5. Keep it moving. Add things as they happen: a talk, a role change, a piece you wrote. Freshness is weighed.
  6. Measure it. Ask each engine the questions a stranger would ask about you, every week, and read the answers. Fix the specific gap the answers show. Then ask again.

Step six is the one almost nobody does, and it is the one that turns AEO from a set of beliefs into a feedback loop. You cannot improve what you do not query.

The engines are already answering questions about you. The only question is whether they are working from a page you own or from scraps.