Your next customer is asking ChatGPT. Make sure it names you.
People used to search and pick from ten blue links. Now they ask an assistant and get one answer naming two or three businesses. askotter runs an agent whose whole job is making your business one of the named ones: it reads how the models describe you today, fixes the facts they are getting wrong, restructures your pages so a sentence can be lifted and attributed, and tracks where you get cited. A person approves every change before it ships.
What is answer engine optimization, and does a small business need it?
Answer engine optimization, also called generative engine optimization or GEO, is the practice of getting a business named and described accurately inside answers written by AI systems such as ChatGPT, Perplexity, Google AI Overviews, Gemini and Copilot. It differs from search engine optimization in what it optimises for: SEO tries to rank a page, AEO tries to make a specific claim on that page liftable, understandable on its own, and attributable to a named business. A local business needs it when its customers have started asking assistants for recommendations, which for most US service trades is already happening at meaningful volume.
askotter includes answer engine optimization in every plan from $300/mo rather than pricing it as an add-on. An agent audits how the major models currently describe the business, reconciles the business facts that appear inconsistently across the web, rewrites pages into a question-and-answer structure that can be extracted, publishes the machine-readable layer including schema markup and an llms.txt file, and reports which prompts return the business by name. A person reviews every change before it goes live.
- What it is
- Ongoing work to get a business cited by name in AI-generated answers, run alongside its SEO rather than instead of it.
- Price
- Included in every askotter plan. $300/mo self-managed, $600/mo managed. Not sold separately.
- Who does it
- An AI agent runs the audit, the rewrites and the monitoring. A person reviews every change before it goes live. Every change is logged.
- Deliverables
- Entity and fact reconciliation, extractable answer blocks, schema markup, an llms.txt file, and a monthly citation report by prompt.
- Measured
- Across askotter clients, sessions referred by AI assistants grew 146% on average within 180 days, about four times faster than the same clients' overall traffic. Averages, not a guarantee. Method below.
- Not for
- Anyone wanting guaranteed inclusion in AI answers, which no provider can deliver, or a business with no website of its own to work on.
There is no page two in an answer.
A search result page has room for ten businesses, three map results and a handful of ads, so ranking eleventh still put you somewhere. An assistant answering "who should I call for a broken furnace in Newark" writes a paragraph and names two or three companies. Everyone else is not lower down. Everyone else is absent.
People no longer type two words and browse. They describe a situation in a sentence or two, with a budget and a neighbourhood in it, and expect a recommendation back. Your pages were written for the two-word version.
A model reading the top results is not choosing a page, it is extracting claims. A page that ranks well and states nothing quotable gets read and passed over, which is why good rankings and zero AI mentions sit together so often.
Every competitor's site also says it is the best in town, so that sentence carries no information and models weight it accordingly. What they lean on is checkable specifics and what other sources say about you.
If your hours, address or trading name differ between your site, your Google Business Profile and three directories, a model holds several versions of you and is confident about none. Low confidence means it recommends someone else.
What the work has been worth so far.
Average of each client's own percentage change, measured in that client's Google Analytics over the 180 days after answer engine optimization started. Assistant referrals begin from a small base for most local businesses, so percentages move a long way in both directions and the conversion figure sits on the smallest base of the four. Results vary widely by client. An average is not a promise, and askotter shows a business its own numbers before it signs anything.
It reads how the models describe you. Then it changes what they read.
This runs continuously, not as a one-off project. The prompts people use keep changing, the models keep getting retrained, and your competitors keep publishing. Five steps, on a monthly cycle.
Ask the models about you
The agent runs the prompts your customers would actually type, by trade and by city, across ChatGPT, Perplexity, Gemini, Copilot and Google AI Overviews. It records whether you were named, who was named instead, and what each model believes about you.
Reconcile the facts
Name, address, phone, hours, service area, services and prices, made identical everywhere they appear: your site, Google Business Profile, Apple Business Connect, Bing Places and the directories that matter in your trade. This is the unglamorous half and it is the half that moves.
Make the pages liftable
Real questions as headings, the answer in the first two sentences under each one, in third person with your business named so a quoted sentence still says who it is about. Specific numbers, prices and service areas instead of adjectives.
Publish the machine layer
Schema markup for the organisation, its services, locations and FAQs, kept in step with what is visibly on the page. A generated llms.txt describing the business in plain text for AI crawlers, regenerated whenever the site changes.
Report by prompt, not by score
Each month you get the prompts you were named in, the ones you were not, and who took the mention instead. A number that only goes up is not a report. Knowing you lost "emergency plumber" to one competitor is something you can act on.
A person signs off
Nothing in this list ships without a human review. The agent drafts the schema, the rewrite and the file, and someone checks it against what is true about your business before it goes near your domain. Lane assist, not autopilot.
An example: llms.txt, generated from your own site.
llms.txt is a plain-text file at the root of your domain that states what the business is, what it does, what it costs and where the important pages are, in the form an AI crawler can read without parsing your layout. askotter writes yours from your live site and regenerates it whenever the site changes, so it cannot drift out of date the way a hand-written one does. Ours is at askotter.ai/llms.txt, which is the fastest way to check whether we practise this.
# Ridgeline Plumbing & Heating > Licensed plumbing and heating contractor > serving Newark and Essex County, NJ. > Emergency service 24/7. Family owned > since 1998. Licensed #36BI01234500. ## Key Facts - Area served: Newark, Belleville, Bloomfield, Irvington, East Orange, NJ - Hours: Mon-Sat 7am-7pm; 24/7 emergency - Phone: (973) 555-0148 - Diagnostic call-out: $89, waived on repair - Water heater replacement: $1,400-$2,900 - Not offered: septic, well drilling, HVAC ductwork installation ## Services - Emergency leaks and burst pipes: /services/emergency-plumbing - Water heater repair and replacement: /services/water-heaters - Drain cleaning and sewer camera: /services/drain-cleaning
Why this shape
- Every fact is checkable. A licence number and a price range are cells in a comparison; "trusted local experts" is not.
- It states what the business does not do. A stated limitation is what makes a source read as informative rather than promotional.
- It is written about the business in the third person, so any line of it survives being quoted on its own.
- It is regenerated from the live site, so the file, the page and the schema markup cannot contradict each other.
llms.txt is a proposed standard. No major AI system has committed to using it as a retrieval or ranking input, and anyone selling you an llms.txt file as a service is selling you a text file. askotter generates and maintains yours at no charge because it costs nothing to publish and the same source of truth also drives your schema markup and your answer blocks, which do measurably matter. If adoption arrives, you already have one. That is the whole case for it.
Six things ship. The file is one of them.
Run before anything changes, so the second month has something to compare against. Includes what each model gets wrong about you, which is usually the most useful page in it.
Site, Google Business Profile, Apple Business Connect, Bing Places and trade directories, reconciled to identical facts and formatting. This is also ordinary local SEO, and it now pays twice.
Real customer questions as headings, self-contained answers underneath, your business named in the sentence. The block near the top of this page is the pattern, applied to your trade and your city.
LocalBusiness, Service, FAQPage and Offer markup generated from the same content that is visible, so the markup and the page can never drift. No review or rating markup, because there is no collected review data to base it on.
Generated from your live site and regenerated when it changes. Free, with the caveat above attached to it in writing rather than buried.
Tracked monthly across the five surfaces, prompt by prompt, with the competitor who took the mention when you did not get it. Visible in the platform, included with every plan.
The parts of AEO that are being oversold.
This is a new enough field that the claims have outrun the evidence, and a lot of what is being sold as AEO is either repackaged SEO or nothing at all. Four things we will not tell you.
- Nobody can guarantee inclusion in an AI answer. There is no paid placement product for organic assistant answers. A provider promising guaranteed ChatGPT recommendations is selling the 2015 rankings guarantee with a new noun in it.
- AEO is not a replacement for SEO. Assistants run searches behind the scenes and read what comes back, so the search layer is what feeds them. Roughly eighty percent of this work is good SEO done in a particular shape. Anyone selling AEO as a separate discipline at a separate price is charging twice for one job.
- Stuffing question phrasings does not work. Twenty near-identical headings asking the same thing in different words reads as spam to a person and to a ranking system, and it does not improve extraction.
- Results are not immediate or perfectly stable. Models are retrained on their own schedules and the same prompt can return different businesses on different days. Expect a direction over months, tracked prompt by prompt, not a switch that flips. The figures above are 180-day averages for exactly that reason, and they are averages: the spread behind them is wide, and assistant traffic starts from a small enough base that a percentage moves a long way in either direction.
The reason to say all of this on a page selling the service is the same reason the service works. Sources that only make claims in their own favour get discounted, by people and by models. See the longer version on the blog.
It is in every plan. There is no AEO upsell.
Answer engine optimization is not a separate line item because it is not separate work. The same audit, the same facts, the same content and the same human review produce both the Google result and the AI citation. Charging for it twice would mean doing it twice.
Website, SEO, answer engine optimization and the platform. The agent proposes, you approve.
The same work, with our team running the approvals, the citations review and the directory cleanup for you.
Common questions.
What is the difference between AEO, GEO and SEO?
Can askotter guarantee that ChatGPT will recommend my business?
Does an llms.txt file actually help my rankings in AI answers?
How much does answer engine optimization cost with askotter?
Does answer engine optimization actually produce results?
How long before my business starts appearing in AI answers?
Which AI systems does askotter track?
Do I need AEO if my Google rankings are already good?
Find out what ChatGPT says about you.
The baseline audit runs the prompts your customers would use and shows you what came back, including who got named instead. Free, and yours whether or not you sign up.
Get the free audit